Public reference · AI economics

Task-to-Model Catalog.

Choose the model around the work. Follow the evidence as it changes.

Provider check 2026-09-2139 tasks29 model price checksTask validation pending

Weekly source checks · Next review due 2026-09-28 · Partial verification. If the due date has passed, treat this snapshot as overdue.

The useful part of staying current

What’s changed

Provider facts, corrections, and the tasks they affect.

See every change

Provider billing conditionxAI changes X Search billing from calls to fetched results
Provider watchlistGemini 3.8 Live enters the realtime watchlist
Provider watchlistGoogle releases Antigravity Agent 09-2026 and schedules the prior preview shutdown

Explore by task

Selected based on available evidence and expected task fit. EDC testing will assess performance against defined quality, cost and response-time criteria.

Content 6 tasks

USD per 1M text tokens. Expand a task for billing conditions and all 34 columns.
Task / Use Case	Candidate model	Input	Output
CON-001Email & communication drafting All columns & evidence Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID CON-001 Task Category Content Task / Use Case Email & communication drafting Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID anthropic-claude-sonnet-5 Recommended Model Claude Sonnet 5 Provider Anthropic Model Tier Mid Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Input $/1M $2 Output $/1M $10 Blended $/1M $4 Cache $/1M $0.2 Context 1M Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Claude Sonnet 5AnthropicRecommended candidate · Validation pending	$2	$10
CON-002Long-form report & document generation All columns & evidence Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID CON-002 Task Category Content Task / Use Case Long-form report & document generation Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID anthropic-claude-opus-5 Recommended Model Claude Opus 5 Provider Anthropic Model Tier Frontier Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Input $/1M $5 Output $/1M $25 Blended $/1M $10 Cache $/1M $0.5 Context 1M Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Claude Opus 5AnthropicRecommended candidate · Validation pending	$5	$25
CON-003Blog posts & marketing copy All columns & evidence OpenAI Standard API short-context rates. Prompts above 272K use 2× input/cache and 1.5× output rates. Batch/Flex are 50% lower where supported; regional processing adds 10%. Endpoints and 1,050,000-token context were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID CON-003 Task Category Content Task / Use Case Blog posts & marketing copy Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID openai-gpt-5-6-terra Recommended Model GPT-5.6 Terra Provider OpenAI Model Tier Mid Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes OpenAI Standard API short-context rates. Prompts above 272K use 2× input/cache and 1.5× output rates. Batch/Flex are 50% lower where supported; regional processing adds 10%. Endpoints and 1,050,000-token context were rechecked. Input $/1M $2 Output $/1M $12 Blended $/1M $4.5 Cache $/1M $0.2 Context 1,050,000 Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	GPT-5.6 TerraOpenAIRecommended candidate · Validation pending	$2	$12
CON-004Summarization — batch, high-volume All columns & evidence Gemini Developer API paid Standard. Output includes thinking tokens. Cache read excludes storage and grounding/tool charges. Batch/Flex and deprecation status were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M. Task_ID CON-004 Task Category Content Task / Use Case Summarization — batch, high-volume Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID google-gemini-3-5-flash-lite Recommended Model Gemini 3.5 Flash-Lite Provider Google Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Gemini Developer API paid Standard. Output includes thinking tokens. Cache read excludes storage and grounding/tool charges. Batch/Flex and deprecation status were rechecked. Input $/1M $0.3 Output $/1M $2.5 Blended $/1M $0.85 Cache $/1M $0.03 Context Current field verification required Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Gemini 3.5 Flash-LiteGoogleRecommended candidate · Validation pending	$0.3	$2.5
CON-005Translation — enterprise scale All columns & evidence Gemini Developer API paid Standard promotional rates through 31 December 2026. Output includes thinking tokens. Cache read excludes storage and grounding/tool charges; Batch/Flex rates were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M. Task_ID CON-005 Task Category Content Task / Use Case Translation — enterprise scale Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID google-gemini-3-7-flash Recommended Model Gemini 3.7 Flash Provider Google Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Gemini Developer API paid Standard promotional rates through 31 December 2026. Output includes thinking tokens. Cache read excludes storage and grounding/tool charges; Batch/Flex rates were rechecked. Input $/1M $0.75 Output $/1M $3.75 Blended $/1M $1.5 Cache $/1M $0.075 Context Current field verification required Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Gemini 3.7 FlashGoogleRecommended candidate · Validation pending	$0.75	$3.75
CON-006Multilingual content at scale All columns & evidence Alibaba Model Studio Singapore international-region rate for prompts up to 32K. Larger prompt bands and other international regions have different rates. Cache field is implicit cache hit; explicit cache uses separate read/write prices. Batch is not supported. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID CON-006 Task Category Content Task / Use Case Multilingual content at scale Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID alibaba-qwen3-7-flash Recommended Model Qwen3.7 Flash Provider Alibaba Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Alibaba Model Studio Singapore international-region rate for prompts up to 32K. Larger prompt bands and other international regions have different rates. Cache field is implicit cache hit; explicit cache uses separate read/write prices. Batch is not supported. Input $/1M $0.03 Output $/1M $0.13 Blended $/1M $0.055 Cache $/1M $0.006 Context 1,000,000 Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Qwen3.7 FlashAlibabaRecommended candidate · Validation pending	$0.03	$0.13

Code 6 tasks

USD per 1M text tokens. Expand a task for billing conditions and all 34 columns.
Task / Use Case	Candidate model	Input	Output
COD-001Code generation & completion All columns & evidence Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID COD-001 Task Category Code Task / Use Case Code generation & completion Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID anthropic-claude-sonnet-5 Recommended Model Claude Sonnet 5 Provider Anthropic Model Tier Mid Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Input $/1M $2 Output $/1M $10 Blended $/1M $4 Cache $/1M $0.2 Context 1M Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Claude Sonnet 5AnthropicRecommended candidate · Validation pending	$2	$10
COD-002Code review & security analysis All columns & evidence Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID COD-002 Task Category Code Task / Use Case Code review & security analysis Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID anthropic-claude-opus-5 Recommended Model Claude Opus 5 Provider Anthropic Model Tier Frontier Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Input $/1M $5 Output $/1M $25 Blended $/1M $10 Cache $/1M $0.5 Context 1M Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Claude Opus 5AnthropicRecommended candidate · Validation pending	$5	$25
COD-003Agentic coding / autonomous PRs All columns & evidence OpenAI Standard API short-context rates. Promotional input/output pricing is stated through at least 21 November 2026. Prompts above 272K use 2× input/cache and 1.5× output; cache writes cost 1.25×. Batch/Flex are 50% lower where supported. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID COD-003 Task Category Code Task / Use Case Agentic coding / autonomous PRs Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID openai-gpt-5-6-sol Recommended Model GPT-5.6 Sol Provider OpenAI Model Tier Frontier Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes OpenAI Standard API short-context rates. Promotional input/output pricing is stated through at least 21 November 2026. Prompts above 272K use 2× input/cache and 1.5× output; cache writes cost 1.25×. Batch/Flex are 50% lower where supported. Input $/1M $4 Output $/1M $20 Blended $/1M $8 Cache $/1M $0.4 Context 1,050,000 Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	GPT-5.6 SolOpenAIRecommended candidate · Validation pending	$4	$20
COD-004Boilerplate & unit test generation All columns & evidence OpenAI Standard API text rates. Regional processing adds 10%. Supported endpoints and 400K context were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M. Task_ID COD-004 Task Category Code Task / Use Case Boilerplate & unit test generation Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID openai-gpt-5-4-mini Recommended Model GPT-5.4 mini Provider OpenAI Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes OpenAI Standard API text rates. Regional processing adds 10%. Supported endpoints and 400K context were rechecked. Input $/1M $0.75 Output $/1M $4.5 Blended $/1M $1.6875 Cache $/1M $0.075 Context Current field verification required Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	GPT-5.4 miniOpenAIRecommended candidate · Validation pending	$0.75	$4.5
COD-005Cost-optimized coding at scale All columns & evidence The official Kimi pricing URL did not expose verifiable API rates on 21 September 2026. Prior price evidence and 7 September aggregate field date are retained; the official Moonshot model repository separately verifies 1,048,576-token context. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID COD-005 Task Category Code Task / Use Case Cost-optimized coding at scale Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID moonshot-ai-kimi-k3 Recommended Model Kimi K3 Provider Moonshot AI Model Tier Frontier Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes The official Kimi pricing URL did not expose verifiable API rates on 21 September 2026. Prior price evidence and 7 September aggregate field date are retained; the official Moonshot model repository separately verifies 1,048,576-token context. Input $/1M $3 Output $/1M $15 Blended $/1M $6 Cache $/1M $0.3 Context 1,048,576 Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Source review required; Validation pending Source URL Provider source Last Validated 2026-09-07 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Kimi K3Moonshot AIRecommended candidate · Validation pending	$3	$15
COD-006Self-hosted IDE code assistant All columns & evidence Combined self-hosted alternatives require a specific version, deployment and infrastructure estimate. The old “Free” label is withheld; weights do not establish a zero operating cost. Verified fields: None in this check. Task_ID COD-006 Task Category Code Task / Use Case Self-hosted IDE code assistant Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID meta-deepseek-llama-4-scout-deepseek-v4 Recommended Model Llama 4 Scout / DeepSeek V4 Provider Meta / DeepSeek Model Tier Self-host Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Combined self-hosted alternatives require a specific version, deployment and infrastructure estimate. The old “Free” label is withheld; weights do not establish a zero operating cost. Input $/1M Current field verification required Output $/1M Current field verification required Blended $/1M Not a comparable complete workload estimate Cache $/1M Current field verification required Context Current field verification required Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Source review required; Validation pending Source URL Not stated / not applicable Last Validated Not stated / not applicable EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Llama 4 Scout / DeepSeek V4Meta / DeepSeekRecommended candidate · Validation pending	—	—

Data 5 tasks

USD per 1M text tokens. Expand a task for billing conditions and all 34 columns.
Task / Use Case	Candidate model	Input	Output
DAT-001Structured data extraction (JSON/CSV) All columns & evidence MiniMax direct API rates apply through 512K input; larger inputs use a higher price band. Priority costs 1.5×. The model has 1M context. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID DAT-001 Task Category Data Task / Use Case Structured data extraction (JSON/CSV) Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID minimax-minimax-m3 Recommended Model MiniMax M3 Provider MiniMax Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes MiniMax direct API rates apply through 512K input; larger inputs use a higher price band. Priority costs 1.5×. The model has 1M context. Input $/1M $0.3 Output $/1M $1.2 Blended $/1M $0.525 Cache $/1M $0.06 Context 1M Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	MiniMax M3MiniMaxRecommended candidate · Validation pending	$0.3	$1.2
DAT-002SQL generation from natural language All columns & evidence xAI direct API rates below the 200K prompt threshold. Prompts at or above 200K cost $4 input / $0.60 cached / $12 output. Priority costs 2×; Batch eligibility and geography/account availability were rechecked. Effective 21 September 2026 at 12:00 p.m. PT, X Search bills $5 per 1,000 posts fetched and $10 per 1,000 profiles fetched instead of per call. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID DAT-002 Task Category Data Task / Use Case SQL generation from natural language Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID xai-grok-4-5 Recommended Model Grok 4.5 Provider xAI Model Tier Mid Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes xAI direct API rates below the 200K prompt threshold. Prompts at or above 200K cost $4 input / $0.60 cached / $12 output. Priority costs 2×; Batch eligibility and geography/account availability were rechecked. Effective 21 September 2026 at 12:00 p.m. PT, X Search bills $5 per 1,000 posts fetched and $10 per 1,000 profiles fetched instead of per call. Input $/1M $2 Output $/1M $6 Blended $/1M $3 Cache $/1M $0.3 Context 500K Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Grok 4.5xAIRecommended candidate · Validation pending	$2	$6
DAT-003Data pipeline orchestration (agentic) All columns & evidence Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID DAT-003 Task Category Data Task / Use Case Data pipeline orchestration (agentic) Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID anthropic-claude-opus-5 Recommended Model Claude Opus 5 Provider Anthropic Model Tier Frontier Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Input $/1M $5 Output $/1M $25 Blended $/1M $10 Cache $/1M $0.5 Context 1M Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Claude Opus 5AnthropicRecommended candidate · Validation pending	$5	$25
DAT-004Embedding generation (vector search) All columns & evidence OpenAI embedding input price and Batch endpoint availability were rechecked. This is not a generative text model, so no output or cache price is asserted. Verified fields: Input $/1M, Output $/1M, Cache $/1M. Task_ID DAT-004 Task Category Data Task / Use Case Embedding generation (vector search) Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID openai-text-embedding-3-small Recommended Model text-embedding-3-small Provider OpenAI Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes OpenAI embedding input price and Batch endpoint availability were rechecked. This is not a generative text model, so no output or cache price is asserted. Input $/1M $0.02 Output $/1M Not stated or not applicable; not zero Blended $/1M Not a comparable complete workload estimate Cache $/1M Not stated or not applicable; not zero Context Current field verification required Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	text-embedding-3-smallOpenAIRecommended candidate · Validation pending	$0.02	—
DAT-005High-volume data processing All columns & evidence The legacy deepseek-v4-flash alias is accepted but now serves DeepSeek-V4.1-Flash. Peak input / output / cache hit are $0.30 / $1.20 / $0.006 per 1M; off-peak rates are half. Peak hours are 01:00–04:00 and 06:00–10:00 UTC Monday–Friday. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context, Serving model. Task_ID DAT-005 Task Category Data Task / Use Case High-volume data processing Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID deepseek-deepseek-v4-flash Recommended Model DeepSeek V4 Flash Provider DeepSeek Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes The legacy deepseek-v4-flash alias is accepted but now serves DeepSeek-V4.1-Flash. Peak input / output / cache hit are $0.30 / $1.20 / $0.006 per 1M; off-peak rates are half. Peak hours are 01:00–04:00 and 06:00–10:00 UTC Monday–Friday. Input $/1M $0.3 Output $/1M $1.2 Blended $/1M $0.525 Cache $/1M $0.006 Context 1M Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	DeepSeek V4 FlashDeepSeekRecommended candidate · Validation pending	$0.3	$1.2

Reasoning 6 tasks

USD per 1M text tokens. Expand a task for billing conditions and all 34 columns.
Task / Use Case	Candidate model	Input	Output
REA-001Legal & contract analysis All columns & evidence Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID REA-001 Task Category Reasoning Task / Use Case Legal & contract analysis Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID anthropic-claude-opus-5 Recommended Model Claude Opus 5 Provider Anthropic Model Tier Frontier Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Input $/1M $5 Output $/1M $25 Blended $/1M $10 Cache $/1M $0.5 Context 1M Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Claude Opus 5AnthropicRecommended candidate · Validation pending	$5	$25
REA-002Financial modeling & scenario analysis All columns & evidence OpenAI Standard API short-context rates. Promotional input/output pricing is stated through at least 21 November 2026. Prompts above 272K use 2× input/cache and 1.5× output; cache writes cost 1.25×. Batch/Flex are 50% lower where supported. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID REA-002 Task Category Reasoning Task / Use Case Financial modeling & scenario analysis Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID openai-gpt-5-6-sol Recommended Model GPT-5.6 Sol Provider OpenAI Model Tier Frontier Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes OpenAI Standard API short-context rates. Promotional input/output pricing is stated through at least 21 November 2026. Prompts above 272K use 2× input/cache and 1.5× output; cache writes cost 1.25×. Batch/Flex are 50% lower where supported. Input $/1M $4 Output $/1M $20 Blended $/1M $8 Cache $/1M $0.4 Context 1,050,000 Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	GPT-5.6 SolOpenAIRecommended candidate · Validation pending	$4	$20
REA-003Research synthesis & literature review All columns & evidence Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID REA-003 Task Category Reasoning Task / Use Case Research synthesis & literature review Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID anthropic-claude-opus-5 Recommended Model Claude Opus 5 Provider Anthropic Model Tier Frontier Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Input $/1M $5 Output $/1M $25 Blended $/1M $10 Cache $/1M $0.5 Context 1M Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Claude Opus 5AnthropicRecommended candidate · Validation pending	$5	$25
REA-004Highest-stakes reasoning (regulated) All columns & evidence Direct Claude API pricing. Mythos 5 remains limited to approved Project Glasswing customers. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID REA-004 Task Category Reasoning Task / Use Case Highest-stakes reasoning (regulated) Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID anthropic-claude-mythos-5 Recommended Model Claude Mythos 5 Provider Anthropic Model Tier Frontier Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Direct Claude API pricing. Mythos 5 remains limited to approved Project Glasswing customers. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. Input $/1M $10 Output $/1M $50 Blended $/1M $20 Cache $/1M $1 Context 1M Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Claude Mythos 5AnthropicRecommended candidate · Validation pending	$10	$50
REA-005Strategic planning & ideation All columns & evidence Gemini Developer API paid Standard promotional rates through 31 December 2026. Output includes thinking tokens. Cache read excludes storage and grounding/tool charges; Batch/Flex rates were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M. Task_ID REA-005 Task Category Reasoning Task / Use Case Strategic planning & ideation Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID google-gemini-3-6-flash Recommended Model Gemini 3.6 Flash Provider Google Model Tier Mid Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Gemini Developer API paid Standard promotional rates through 31 December 2026. Output includes thinking tokens. Cache read excludes storage and grounding/tool charges; Batch/Flex rates were rechecked. Input $/1M $0.75 Output $/1M $3.75 Blended $/1M $1.5 Cache $/1M $0.075 Context Current field verification required Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Gemini 3.6 FlashGoogleRecommended candidate · Validation pending	$0.75	$3.75
REA-006Budget reasoning at scale All columns & evidence DeepSeek states that V4 Pro API service continues after 14 September 2026 with billing unchanged. Peak input / output / cache hit are $1.32 / $3.96 / $0.044 per 1M; off-peak rates are half. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context, Availability status. Task_ID REA-006 Task Category Reasoning Task / Use Case Budget reasoning at scale Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID deepseek-deepseek-v4-pro Recommended Model DeepSeek V4 Pro Provider DeepSeek Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes DeepSeek states that V4 Pro API service continues after 14 September 2026 with billing unchanged. Peak input / output / cache hit are $1.32 / $3.96 / $0.044 per 1M; off-peak rates are half. Input $/1M $1.32 Output $/1M $3.96 Blended $/1M $1.98 Cache $/1M $0.044 Context 1M Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	DeepSeek V4 ProDeepSeekRecommended candidate · Validation pending	$1.32	$3.96

Agentic 5 tasks

USD per 1M text tokens. Expand a task for billing conditions and all 34 columns.
Task / Use Case	Candidate model	Input	Output
AGE-001Multi-step workflow automation All columns & evidence Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID AGE-001 Task Category Agentic Task / Use Case Multi-step workflow automation Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID anthropic-claude-opus-5 Recommended Model Claude Opus 5 Provider Anthropic Model Tier Frontier Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. US-only inference costs 1.1× base rates. 1M context and active model status were rechecked. Input $/1M $5 Output $/1M $25 Blended $/1M $10 Cache $/1M $0.5 Context 1M Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Claude Opus 5AnthropicRecommended candidate · Validation pending	$5	$25
AGE-002Customer support agent (high-volume) All columns & evidence Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. 200K context and active model status were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M. Task_ID AGE-002 Task Category Agentic Task / Use Case Customer support agent (high-volume) Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID anthropic-claude-haiku-4-5 Recommended Model Claude Haiku 4.5 Provider Anthropic Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Direct Claude API base rates. Cache price is a cache read; writes cost extra. Batch token rates are 50% lower. 200K context and active model status were rechecked. Input $/1M $1 Output $/1M $5 Blended $/1M $2 Cache $/1M $0.1 Context Current field verification required Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Claude Haiku 4.5AnthropicRecommended candidate · Validation pending	$1	$5
AGE-003Internal IT helpdesk automation All columns & evidence Gemini Developer API paid Standard. No shutdown date is announced, but Google names Gemini 3.6 Flash as the replacement. Cache storage and grounding/tool charges are additional. Candidate mapping remains unchanged pending evaluation. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Recommended Model, Provider replacement. Task_ID AGE-003 Task Category Agentic Task / Use Case Internal IT helpdesk automation Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID google-gemini-3-flash Recommended Model Gemini 3 Flash Preview Provider Google Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Gemini Developer API paid Standard. No shutdown date is announced, but Google names Gemini 3.6 Flash as the replacement. Cache storage and grounding/tool charges are additional. Candidate mapping remains unchanged pending evaluation. Input $/1M $0.5 Output $/1M $3 Blended $/1M $1.125 Cache $/1M $0.05 Context Current field verification required Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Gemini 3 Flash PreviewGoogleRecommended candidate · Validation pending	$0.5	$3
AGE-004Code execution & DevOps agent All columns & evidence OpenAI Standard API short-context rates. Prompts above 272K use 2× input/cache and 1.5× output rates. Batch/Flex are 50% lower where supported; regional processing adds 10%. Endpoints and 1,050,000-token context were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID AGE-004 Task Category Agentic Task / Use Case Code execution & DevOps agent Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID openai-gpt-5-6-terra Recommended Model GPT-5.6 Terra Provider OpenAI Model Tier Mid Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes OpenAI Standard API short-context rates. Prompts above 272K use 2× input/cache and 1.5× output rates. Batch/Flex are 50% lower where supported; regional processing adds 10%. Endpoints and 1,050,000-token context were rechecked. Input $/1M $2 Output $/1M $12 Blended $/1M $4.5 Cache $/1M $0.2 Context 1,050,000 Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	GPT-5.6 TerraOpenAIRecommended candidate · Validation pending	$2	$12
AGE-005Cost-sensitive agent orchestration All columns & evidence xAI direct API rates below the 200K prompt threshold. Prompts at or above 200K cost $4 input / $1 cached / $12 output. Priority costs 2×; Batch eligibility and geography/account availability were rechecked. Effective 21 September 2026 at 12:00 p.m. PT, X Search bills $5 per 1,000 posts fetched and $10 per 1,000 profiles fetched instead of per call. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID AGE-005 Task Category Agentic Task / Use Case Cost-sensitive agent orchestration Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID xai-grok-4-6 Recommended Model Grok 4.6 Provider xAI Model Tier Mid Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes xAI direct API rates below the 200K prompt threshold. Prompts at or above 200K cost $4 input / $1 cached / $12 output. Priority costs 2×; Batch eligibility and geography/account availability were rechecked. Effective 21 September 2026 at 12:00 p.m. PT, X Search bills $5 per 1,000 posts fetched and $10 per 1,000 profiles fetched instead of per call. Input $/1M $2 Output $/1M $6 Blended $/1M $3 Cache $/1M $0.5 Context 500K Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Grok 4.6xAIRecommended candidate · Validation pending	$2	$6

Vision 4 tasks

USD per 1M text tokens. Expand a task for billing conditions and all 34 columns.
Task / Use Case	Candidate model	Input	Output
VIS-001Document & image OCR extraction All columns & evidence Gemini Developer API paid Standard. Output includes thinking tokens. Cache read excludes storage and grounding/tool charges. Context-band, Batch/Flex and deprecation conditions were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M. Task_ID VIS-001 Task Category Vision Task / Use Case Document & image OCR extraction Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID google-gemini-3-5-flash Recommended Model Gemini 3.5 Flash Provider Google Model Tier Mid Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Gemini Developer API paid Standard. Output includes thinking tokens. Cache read excludes storage and grounding/tool charges. Context-band, Batch/Flex and deprecation conditions were rechecked. Input $/1M $1.5 Output $/1M $9 Blended $/1M $3.375 Cache $/1M $0.15 Context Current field verification required Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Gemini 3.5 FlashGoogleRecommended candidate · Validation pending	$1.5	$9
VIS-002Video analysis & summarization All columns & evidence Gemini Developer API paid Standard preview. Base rates apply through 200K prompt tokens; longer prompts use higher rates. 1,048,576 input / 65,536 output context and preview status were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M. Task_ID VIS-002 Task Category Vision Task / Use Case Video analysis & summarization Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID google-gemini-3-1-pro-preview Recommended Model Gemini 3.1 Pro Preview Provider Google Model Tier Frontier Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Gemini Developer API paid Standard preview. Base rates apply through 200K prompt tokens; longer prompts use higher rates. 1,048,576 input / 65,536 output context and preview status were rechecked. Input $/1M $2 Output $/1M $12 Blended $/1M $4.5 Cache $/1M $0.2 Context Current field verification required Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Gemini 3.1 Pro PreviewGoogleRecommended candidate · Validation pending	$2	$12
VIS-003Chart & diagram interpretation All columns & evidence OpenAI Standard API short-context rates. Prompts above 272K use 2× input/cache and 1.5× output rates. Batch/Flex are 50% lower where supported; regional processing adds 10%. Endpoints and 1,050,000-token context were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID VIS-003 Task Category Vision Task / Use Case Chart & diagram interpretation Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID openai-gpt-5-6-terra Recommended Model GPT-5.6 Terra Provider OpenAI Model Tier Mid Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes OpenAI Standard API short-context rates. Prompts above 272K use 2× input/cache and 1.5× output rates. Batch/Flex are 50% lower where supported; regional processing adds 10%. Endpoints and 1,050,000-token context were rechecked. Input $/1M $2 Output $/1M $12 Blended $/1M $4.5 Cache $/1M $0.2 Context 1,050,000 Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	GPT-5.6 TerraOpenAIRecommended candidate · Validation pending	$2	$12
VIS-004Computer use / UI automation All columns & evidence H Company direct API rates for holo3-1-35b-a3b. Cache pricing is not stated; a missing price is not zero. Paid/free access, 65,536 context and zero-data-retention statement were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID VIS-004 Task Category Vision Task / Use Case Computer use / UI automation Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID h-company-holo3-1-35b-a3b Recommended Model Holo3.1-35B-A3B Provider H Company Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes H Company direct API rates for holo3-1-35b-a3b. Cache pricing is not stated; a missing price is not zero. Paid/free access, 65,536 context and zero-data-retention statement were rechecked. Input $/1M $0.25 Output $/1M $1.8 Blended $/1M $0.6375 Cache $/1M Not stated or not applicable; not zero Context 65,536 Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Holo3.1-35B-A3BH CompanyRecommended candidate · Validation pending	$0.25	$1.8

Classification 5 tasks

USD per 1M text tokens. Expand a task for billing conditions and all 34 columns.
Task / Use Case	Candidate model	Input	Output
CLA-001Sentiment & intent classification All columns & evidence OpenAI Standard API short-context rates. Prompts above 272K use 2× input/cache and 1.5× output. Batch/Flex are 50% lower where supported; regional processing adds 10%. Endpoints and 1,050,000-token context were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID CLA-001 Task Category Classification Task / Use Case Sentiment & intent classification Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID openai-gpt-5-6-luna Recommended Model GPT-5.6 Luna Provider OpenAI Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes OpenAI Standard API short-context rates. Prompts above 272K use 2× input/cache and 1.5× output. Batch/Flex are 50% lower where supported; regional processing adds 10%. Endpoints and 1,050,000-token context were rechecked. Input $/1M $0.2 Output $/1M $1.2 Blended $/1M $0.45 Cache $/1M $0.02 Context 1,050,000 Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	GPT-5.6 LunaOpenAIRecommended candidate · Validation pending	$0.2	$1.2
CLA-002Content moderation All columns & evidence Gemini Developer API paid Standard. Google schedules shutdown for 7 May 2027 and names Gemini 3.5 Flash-Lite as replacement. Cache storage and grounding/tool charges are additional. Candidate mapping remains unchanged pending evaluation. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Deprecation status. Task_ID CLA-002 Task Category Classification Task / Use Case Content moderation Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID google-gemini-3-1-flash-lite Recommended Model Gemini 3.1 Flash-Lite Provider Google Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Gemini Developer API paid Standard. Google schedules shutdown for 7 May 2027 and names Gemini 3.5 Flash-Lite as replacement. Cache storage and grounding/tool charges are additional. Candidate mapping remains unchanged pending evaluation. Input $/1M $0.25 Output $/1M $1.5 Blended $/1M $0.5625 Cache $/1M $0.025 Context Current field verification required Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Gemini 3.1 Flash-LiteGoogleRecommended candidate · Validation pending	$0.25	$1.5
CLA-003Entity extraction / NER All columns & evidence Mistral direct inference API token and cached-input rates. 256K context, GA status and Batch endpoint support were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID CLA-003 Task Category Classification Task / Use Case Entity extraction / NER Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID mistral-mistral-small-4 Recommended Model Mistral Small 4 Provider Mistral Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Mistral direct inference API token and cached-input rates. 256K context, GA status and Batch endpoint support were rechecked. Input $/1M $0.15 Output $/1M $0.6 Blended $/1M $0.2625 Cache $/1M $0.015 Context 256K Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Mistral Small 4MistralRecommended candidate · Validation pending	$0.15	$0.6
CLA-004Routing / triage (agentic orchestration) All columns & evidence Mistral direct inference API token and cached-input rates. 256K context, GA status and Batch endpoint support were rechecked; self-hosting has separate infrastructure cost. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID CLA-004 Task Category Classification Task / Use Case Routing / triage (agentic orchestration) Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID mistral-ministral-3-3b Recommended Model Ministral 3 3B Provider Mistral Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Mistral direct inference API token and cached-input rates. 256K context, GA status and Batch endpoint support were rechecked; self-hosting has separate infrastructure cost. Input $/1M $0.1 Output $/1M $0.1 Blended $/1M $0.1 Cache $/1M $0.01 Context 256K Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Ministral 3 3BMistralRecommended candidate · Validation pending	$0.1	$0.1
CLA-005Ultra-high-volume triage All columns & evidence StepFun direct API cache-miss, cache-hit and output rates. Active listing and account-tier rate limits were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M. Task_ID CLA-005 Task Category Classification Task / Use Case Ultra-high-volume triage Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID stepfun-step-3-5-flash Recommended Model Step 3.5 Flash Provider StepFun Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes StepFun direct API cache-miss, cache-hit and output rates. Active listing and account-tier rate limits were rechecked. Input $/1M $0.1 Output $/1M $0.3 Blended $/1M $0.15 Cache $/1M $0.02 Context Current field verification required Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	Step 3.5 FlashStepFunRecommended candidate · Validation pending	$0.1	$0.3

Realtime 2 tasks

USD per 1M text tokens. Expand a task for billing conditions and all 34 columns.
Task / Use Case	Candidate model	Input	Output
REA-007Voice agent / realtime conversation All columns & evidence TEXT tokens only: $4 input / $24 output / $0.40 cached input. Audio is $32 / $64 / $0.40 per 1M; image tokens have separate rates. Exact-model context and realtime serving endpoints were rechecked. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context. Task_ID REA-007 Task Category Realtime Task / Use Case Voice agent / realtime conversation Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID openai-gpt-realtime-2 Recommended Model GPT Realtime 2 Provider OpenAI Model Tier Frontier Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes TEXT tokens only: $4 input / $24 output / $0.40 cached input. Audio is $32 / $64 / $0.40 per 1M; image tokens have separate rates. Exact-model context and realtime serving endpoints were rechecked. Input $/1M $4 Output $/1M $24 Blended $/1M Not a comparable complete workload estimate Cache $/1M $0.4 Context 128,000 Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Provider fields checked; Validation pending Source URL Provider source Last Validated 2026-09-21 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	GPT Realtime 2OpenAIRecommended candidate · Validation pending	$4	$24
REA-008Cost-optimized voice agent All columns & evidence Text rates, 32,000-token context and the 20 January 2027 alias shutdown were rechecked on 21 September 2026. Exact-model audio pricing was not available on the model page, so the aggregate checked_on date remains 7 September 2026 and audio pricing is withheld. Verified fields: Input $/1M, Output $/1M, Cache $/1M, Context, Deprecation status. Task_ID REA-008 Task Category Realtime Task / Use Case Cost-optimized voice agent Job Function All Seniority Min Entry Seniority Max Executive Risk Tier Pending policy/claim review Data Classification Allowed Pending policy/claim review Model_ID openai-gpt-realtime-mini Recommended Model GPT Realtime mini Provider OpenAI Model Tier Budget Access Status Pending policy/claim review Recommendation Rank Task comparison not completed Quality Score Validation pending Cost Score Validation pending Latency Score Validation pending Fallback Model Fallback evaluation required Approval Required Pending policy/claim review Best Use When Pending policy/claim review Avoid Pending policy/claim review Optimization Notes Text rates, 32,000-token context and the 20 January 2027 alias shutdown were rechecked on 21 September 2026. Exact-model audio pricing was not available on the model page, so the aggregate checked_on date remains 7 September 2026 and audio pricing is withheld. Input $/1M $0.6 Output $/1M $2.4 Blended $/1M Not a comparable complete workload estimate Cache $/1M $0.06 Context 32,000 Effective From Effective date not established by this check Effective To Effective date not established by this check Validation Status Source review required; Validation pending Source URL Provider source Last Validated 2026-09-07 EDC perspective Pending EDC editorial review EDC next action Pending EDC editorial review	GPT Realtime miniOpenAIRecommended candidate · Validation pending	$0.6	$2.4

Read the evidence, then test the task.

Original column names and order are preserved, including “Recommended Model.” On this page that field identifies a recommended candidate based on available evidence and expected task fit. Task validation assesses that recommendation against documented criteria. “Blended $/1M” assumes 3:1 input to output; it is not cost per completed task.

EDC perspective and EDC next action are pending editorial review. Unverified context values, comparative advice, and policy permissions are blank in public exports with an explicit reason in the JSON metadata. Missing prices never mean zero.

Open verification items

Kimi K3 API pricing could not be reverified because the official pricing URL did not expose API rates; prior values and 7 September aggregate date were retained while context was independently rechecked.
Realtime Mini audio pricing still needs an exact-model primary source; successful text/context/deprecation checks did not advance the aggregate field date.
Self-hosted composite row needs a specific model version, deployment region, infrastructure basis and matching cost source.
Context values not listed in verified model_evidence fields remain withheld from the public projection.

Follow or reuse

Connect to the catalog.

Paste the RSS URL into your reader. Use the JSON feed for change notifications or the catalog snapshot for all 34 columns. No account is required.

Subscribe via RSS JSON Feed Catalog JSON Download CSV

https://ericdriscollconsulting.com/task-to-model/feed.xml

Stable task and change IDs · Schema 1.0 · Read-only public files · Weekly review target

For integrations, fetch from your server or feed reader. Check next_review_due before using a snapshot; check HTTP cache validators before downloading again. Changed facts get a permanent item; a check with no material change does not create a feed notification. Cite EDC, the revision, and the linked provider source when reusing an entry.

Need to turn candidates into a routing decision? Work with Eric on the task, evaluation, and economics.