Loading…

Research & analyze

Research a topic

Refine candidates (optional)

Advanced requirements

Official API requiredMust run locallyMust work entirely offline

Candidates to compare

Based on documented capabilities, not measured performance on your task. Leaderboard context is shown separately.

What to test first

Give every candidate the same source packet. Ask for a sourced summary and conflicting evidence; open the cited passages and count unsupported claims.

GPT-5.5

OpenAI · developers.openai.com

LMArena #28 ↗gpt-5.5 · Text · Style control · Overall2026-09-13

Ranked configurations · 1

Ranked configurationsLMArena #20 · High ↗2026-09-13 · Not the current configuration’s rank

Source and configuration

These ranks belong to the listed evaluation configurations, not the current configuration or measured quality on your task.

gpt-5.5-high · Text · Style control · OverallSource ↗

snapshot:8361fb781955a675a1d6766d3baafd383abd211caf1b589972affe35c9368f0a

Lower recorded token rates than GPT-6 Astra; compare task quality in a trial.

Context window[1]
1,050,000 tokens
Cost[2]
USD 5 input / 30 output per 1M tokens; standard tier, additional charges may apply
Standard ≤272K input tier; above that input ×2, output ×1.5. Caching/tools separate.

Full evidence and sources

LMArena #28 ↗gpt-5.5 · Text · Style control · Overall2026-09-13

Ranked configurations · 1

Ranked configurationsLMArena #20 · High ↗2026-09-13 · Not the current configuration’s rank

Source and configuration

These ranks belong to the listed evaluation configurations, not the current configuration or measured quality on your task.

gpt-5.5-high · Text · Style control · OverallSource ↗

snapshot:8361fb781955a675a1d6766d3baafd383abd211caf1b589972affe35c9368f0a

Why this fits

Both recorded token rates are no higher than GPT-6 Astra (input 10, output 50 USD per 1M tokens), and at least one is lower. Consider it for repeated trials; lower rates do not prove equal quality.

GPT-5.5 · 2026-09-30T17:37:41.000Z · View source GPT-6 Astra · 2026-09-30T17:37:41.000Z · View source

Context window
1050000 · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Official feature documentation; no same-task quality measurement.
This is the documented context capacity, useful when testing a shared source packet across several documents. It does not guarantee retrieval accuracy, reliable citations, or live web access.
Cost
USD 5 input / 30 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source
Standard ≤272K input tier; above that input ×2, output ×1.5. Caching/tools separate.
Use the recorded price as a budget input, not a total project quote. Usage, plan entitlements, taxes, and extra services can change the actual cost.

Strengths

Cost:
USD 5 input / 30 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Standard ≤272K input tier; above that input ×2, output ×1.5. Caching/tools separate.

Trade-offs

Unreviewed dimensions remain open questions for your trial.

Cost: USD 5 input / 30 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Standard ≤272K input tier; above that input ×2, output ×1.5. Caching/tools separate.
Plan-dependent. Check the official price and billing period.
Access: Local deployment:
Unknown
Official API:
Supported · Official fact · Current · View source
Workflow: Use the official API, then import the output into Proof Lab.
Commercial use and rights: Unknown
Rights depend on the product, plan, inputs and intended use. Review the linked terms.

Unknowns

Correctness · Source quality · Explanation quality

Evidence

Task support

Supported · Editorial assessment · Current · View source Last verified: 2026-09-30T17:29:41ZConfidence: Low

Context window

1050000 · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Official feature documentation; no same-task quality measurement.

Tool calling

Supported · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Official feature documentation; no same-task quality measurement.

Local deployment

UnknownLast verified: Not reviewedConfidence: Low

Official API

Supported · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Publicly documented channel; account, region, entitlement and quota were not tested.

Cost

USD 5 input / 30 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Standard ≤272K input tier; above that input ×2, output ×1.5. Caching/tools separate.

Commercial use

UnknownLast verified: Not reviewedConfidence: Low

Claude Fable 5.1

Anthropic · platform.claude.com

Ranked configurationsLMArena #5 · Max ↗2026-09-13 · Not the current configuration’s rank

Source and configuration

These ranks belong to the listed evaluation configurations, not the current configuration or measured quality on your task.

claude-fable-5.1-max · Text · Style control · OverallSource ↗

snapshot:8361fb781955a675a1d6766d3baafd383abd211caf1b589972affe35c9368f0a

Higher recorded token rates than Claude Opus 5; test whether the result justifies the cost.

Context window[1]
1,000,000 tokens
Cost[2]
USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply
Standard API rate; cache writes/reads and tools separate. Not a consumer subscription or total project quote.

Full evidence and sources

Ranked configurationsLMArena #5 · Max ↗2026-09-13 · Not the current configuration’s rank

Source and configuration

These ranks belong to the listed evaluation configurations, not the current configuration or measured quality on your task.

claude-fable-5.1-max · Text · Style control · OverallSource ↗

snapshot:8361fb781955a675a1d6766d3baafd383abd211caf1b589972affe35c9368f0a

Why this fits

Both recorded token rates are no lower than Claude Opus 5 (input 5, output 25 USD per 1M tokens), and at least one is higher. Keep it only if your trial justifies the additional spend.

Claude Fable 5.1 · 2026-09-30T17:37:41.000Z · View source Claude Opus 5 · 2026-09-30T17:37:41.000Z · View source

Context window
1000000 · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Official feature documentation; no same-task quality measurement.
This is the documented context capacity, useful when testing a shared source packet across several documents. It does not guarantee retrieval accuracy, reliable citations, or live web access.
Cost
USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source
Standard API rate; cache writes/reads and tools separate. Not a consumer subscription or total project quote.
Use the recorded price as a budget input, not a total project quote. Usage, plan entitlements, taxes, and extra services can change the actual cost.

Strengths

Cost:
USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Standard API rate; cache writes/reads and tools separate. Not a consumer subscription or total project quote.

Trade-offs

Unreviewed dimensions remain open questions for your trial.

Cost: USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Standard API rate; cache writes/reads and tools separate. Not a consumer subscription or total project quote.
Plan-dependent. Check the official price and billing period.
Access: Local deployment:
Unknown
Official API:
Supported · Official fact · Current · View source
Workflow: Use the official API, then import the output into Proof Lab.
Commercial use and rights: Unknown
Rights depend on the product, plan, inputs and intended use. Review the linked terms.

Unknowns

Correctness · Source quality · Explanation quality

Evidence

Task support

Supported · Editorial assessment · Current · View source Last verified: 2026-09-30T17:29:41ZConfidence: Low

Context window

1000000 · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Official feature documentation; no same-task quality measurement.

Tool calling

Supported · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Official feature documentation; no same-task quality measurement.

Local deployment

UnknownLast verified: Not reviewedConfidence: Low

Official API

Supported · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Publicly documented channel; account, region, entitlement and quota were not tested.

Cost

USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Standard API rate; cache writes/reads and tools separate. Not a consumer subscription or total project quote.

Commercial use

UnknownLast verified: Not reviewedConfidence: Low

Claude Opus 5

Anthropic · platform.claude.com

Ranked configurationsLMArena #10 · High ↗LMArena #14 · Max ↗2026-09-13 · Not the current configuration’s rank

Source and configuration

These ranks belong to the listed evaluation configurations, not the current configuration or measured quality on your task.

claude-opus-5-high · Text · Style control · OverallSource ↗

claude-opus-5-max · Text · Style control · OverallSource ↗

snapshot:8361fb781955a675a1d6766d3baafd383abd211caf1b589972affe35c9368f0a

Lower recorded token rates than Claude Fable 5.1; compare task quality in a trial.

Context window[1]
1,000,000 tokens
Cost[2]
USD 5 input / 25 output per 1M tokens; standard tier, additional charges may apply
Standard API rate; cache writes/reads and tools separate. Not a consumer subscription or total project quote.

Full evidence and sources

Ranked configurationsLMArena #10 · High ↗LMArena #14 · Max ↗2026-09-13 · Not the current configuration’s rank

Source and configuration

These ranks belong to the listed evaluation configurations, not the current configuration or measured quality on your task.

claude-opus-5-high · Text · Style control · OverallSource ↗

claude-opus-5-max · Text · Style control · OverallSource ↗

snapshot:8361fb781955a675a1d6766d3baafd383abd211caf1b589972affe35c9368f0a

Why this fits

Both recorded token rates are no higher than Claude Fable 5.1 (input 10, output 50 USD per 1M tokens), and at least one is lower. Consider it for repeated trials; lower rates do not prove equal quality.

Claude Opus 5 · 2026-09-30T17:37:41.000Z · View source Claude Fable 5.1 · 2026-09-30T17:37:41.000Z · View source

Context window
1000000 · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Official feature documentation; no same-task quality measurement.
This is the documented context capacity, useful when testing a shared source packet across several documents. It does not guarantee retrieval accuracy, reliable citations, or live web access.
Cost
USD 5 input / 25 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source
Standard API rate; cache writes/reads and tools separate. Not a consumer subscription or total project quote.
Use the recorded price as a budget input, not a total project quote. Usage, plan entitlements, taxes, and extra services can change the actual cost.

Strengths

Cost:
USD 5 input / 25 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Standard API rate; cache writes/reads and tools separate. Not a consumer subscription or total project quote.

Trade-offs

Unreviewed dimensions remain open questions for your trial.

Cost: USD 5 input / 25 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Standard API rate; cache writes/reads and tools separate. Not a consumer subscription or total project quote.
Plan-dependent. Check the official price and billing period.
Access: Local deployment:
Unknown
Official API:
Supported · Official fact · Current · View source
Workflow: Use the official API, then import the output into Proof Lab.
Commercial use and rights: Unknown
Rights depend on the product, plan, inputs and intended use. Review the linked terms.

Unknowns

Correctness · Source quality · Explanation quality

Evidence

Task support

Supported · Editorial assessment · Current · View source Last verified: 2026-09-30T17:29:41ZConfidence: Low

Context window

1000000 · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Official feature documentation; no same-task quality measurement.

Tool calling

Supported · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Official feature documentation; no same-task quality measurement.

Local deployment

UnknownLast verified: Not reviewedConfidence: Low

Official API

Supported · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Publicly documented channel; account, region, entitlement and quota were not tested.

Cost

USD 5 input / 25 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Standard API rate; cache writes/reads and tools separate. Not a consumer subscription or total project quote.

Commercial use

UnknownLast verified: Not reviewedConfidence: Low

Gemini 3.7 Flash

Google · ai.google.dev

Ranked configurationsLMArena #12 · High ↗2026-09-13 · Not the current configuration’s rank

Source and configuration

These ranks belong to the listed evaluation configurations, not the current configuration or measured quality on your task.

gemini-3.7-flash-high · Text · Style control · OverallSource ↗

snapshot:8361fb781955a675a1d6766d3baafd383abd211caf1b589972affe35c9368f0a

Same recorded token rates and context capacity as Gemini 3.8 Flash.

Context window[1]
1,048,576 tokens
Cost[2]
USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply
Standard current rate through 2026-12-31; from 2027-01-01 USD 1.50 input / 7.50 output. Caching and tools separate.

Full evidence and sources

Ranked configurationsLMArena #12 · High ↗2026-09-13 · Not the current configuration’s rank

Source and configuration

These ranks belong to the listed evaluation configurations, not the current configuration or measured quality on your task.

gemini-3.7-flash-high · Text · Style control · OverallSource ↗

snapshot:8361fb781955a675a1d6766d3baafd383abd211caf1b589972affe35c9368f0a

Why this fits

The recorded token rates and context capacity match Gemini 3.8 Flash. These facts do not establish that either version is better; test both on the same material.

Gemini 3.7 Flash · 2026-09-30T17:37:41.000Z · View source Gemini 3.8 Flash · 2026-09-30T17:37:41.000Z · View source

Context window
1048576 · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Official feature documentation; no same-task quality measurement.
This is the documented context capacity, useful when testing a shared source packet across several documents. It does not guarantee retrieval accuracy, reliable citations, or live web access.
Cost
USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source
Standard current rate through 2026-12-31; from 2027-01-01 USD 1.50 input / 7.50 output. Caching and tools separate.
Use the recorded price as a budget input, not a total project quote. Usage, plan entitlements, taxes, and extra services can change the actual cost.

Strengths

Cost:
USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Standard current rate through 2026-12-31; from 2027-01-01 USD 1.50 input / 7.50 output. Caching and tools separate.

Trade-offs

Unreviewed dimensions remain open questions for your trial.

Cost: USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Standard current rate through 2026-12-31; from 2027-01-01 USD 1.50 input / 7.50 output. Caching and tools separate.
Plan-dependent. Check the official price and billing period.
Access: Local deployment:
Unknown
Official API:
Supported · Official fact · Current · View source
Workflow: Use the official API, then import the output into Proof Lab.
Commercial use and rights: Unknown
Rights depend on the product, plan, inputs and intended use. Review the linked terms.

Unknowns

Correctness · Source quality · Explanation quality

Evidence

Task support

Supported · Editorial assessment · Current · View source Last verified: 2026-09-30T17:29:41ZConfidence: Low

Context window

1048576 · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Official feature documentation; no same-task quality measurement.

Tool calling

Supported · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Official feature documentation; no same-task quality measurement.

Local deployment

UnknownLast verified: Not reviewedConfidence: Low

Official API

Supported · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Publicly documented channel; account, region, entitlement and quota were not tested.

Cost

USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Standard current rate through 2026-12-31; from 2027-01-01 USD 1.50 input / 7.50 output. Caching and tools separate.

Commercial use

UnknownLast verified: Not reviewedConfidence: Low

Gemini 3.8 Flash

Google · ai.google.dev

Ranked configurationsLMArena #9 · High ↗2026-09-13 · Not the current configuration’s rank

Source and configuration

These ranks belong to the listed evaluation configurations, not the current configuration or measured quality on your task.

gemini-3.8-flash-high · Text · Style control · OverallSource ↗

snapshot:8361fb781955a675a1d6766d3baafd383abd211caf1b589972affe35c9368f0a

Same recorded token rates and context capacity as Gemini 3.7 Flash.

Context window[1]
1,048,576 tokens
Cost[2]
USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply
Standard current rate through 2026-12-31; from 2027-01-01 USD 1.50 input / 7.50 output. Caching and tools separate.

Full evidence and sources

Ranked configurationsLMArena #9 · High ↗2026-09-13 · Not the current configuration’s rank

Source and configuration

These ranks belong to the listed evaluation configurations, not the current configuration or measured quality on your task.

gemini-3.8-flash-high · Text · Style control · OverallSource ↗

snapshot:8361fb781955a675a1d6766d3baafd383abd211caf1b589972affe35c9368f0a

Why this fits

The recorded token rates and context capacity match Gemini 3.7 Flash. These facts do not establish that either version is better; test both on the same material.

Gemini 3.8 Flash · 2026-09-30T17:37:41.000Z · View source Gemini 3.7 Flash · 2026-09-30T17:37:41.000Z · View source

Context window
1048576 · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Official feature documentation; no same-task quality measurement.
This is the documented context capacity, useful when testing a shared source packet across several documents. It does not guarantee retrieval accuracy, reliable citations, or live web access.
Cost
USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source
Standard current rate through 2026-12-31; from 2027-01-01 USD 1.50 input / 7.50 output. Caching and tools separate.
Use the recorded price as a budget input, not a total project quote. Usage, plan entitlements, taxes, and extra services can change the actual cost.

Strengths

Cost:
USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Standard current rate through 2026-12-31; from 2027-01-01 USD 1.50 input / 7.50 output. Caching and tools separate.

Trade-offs

Unreviewed dimensions remain open questions for your trial.

Cost: USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Standard current rate through 2026-12-31; from 2027-01-01 USD 1.50 input / 7.50 output. Caching and tools separate.
Plan-dependent. Check the official price and billing period.
Access: Local deployment:
Unknown
Official API:
Supported · Official fact · Current · View source
Workflow: Use the official API, then import the output into Proof Lab.
Commercial use and rights: Unknown
Rights depend on the product, plan, inputs and intended use. Review the linked terms.

Unknowns

Correctness · Source quality · Explanation quality

Evidence

Task support

Supported · Editorial assessment · Current · View source Last verified: 2026-09-30T17:29:41ZConfidence: Low

Context window

1048576 · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Official feature documentation; no same-task quality measurement.

Tool calling

Supported · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Official feature documentation; no same-task quality measurement.

Local deployment

UnknownLast verified: Not reviewedConfidence: Low

Official API

Supported · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Publicly documented channel; account, region, entitlement and quota were not tested.

Cost

USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Standard current rate through 2026-12-31; from 2027-01-01 USD 1.50 input / 7.50 output. Caching and tools separate.

Commercial use

UnknownLast verified: Not reviewedConfidence: Low

GPT-6 Astra

OpenAI · developers.openai.com

Ranked configurationsLMArena #24 · Max ↗2026-09-13 · Not the current configuration’s rank

Source and configuration

These ranks belong to the listed evaluation configurations, not the current configuration or measured quality on your task.

gpt-6-astra-max · Text · Style control · OverallSource ↗

snapshot:8361fb781955a675a1d6766d3baafd383abd211caf1b589972affe35c9368f0a

Higher recorded token rates than GPT-5.5; test whether the result justifies the cost.

Context window[1]
1,050,000 tokens
Cost[2]
USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply
Standard ≤272K input tier; above that input ×2, output ×1.5. Caching/tools separate.

Full evidence and sources

Ranked configurationsLMArena #24 · Max ↗2026-09-13 · Not the current configuration’s rank

Source and configuration

These ranks belong to the listed evaluation configurations, not the current configuration or measured quality on your task.

gpt-6-astra-max · Text · Style control · OverallSource ↗

snapshot:8361fb781955a675a1d6766d3baafd383abd211caf1b589972affe35c9368f0a

Why this fits

Both recorded token rates are no lower than GPT-5.5 (input 5, output 30 USD per 1M tokens), and at least one is higher. Keep it only if your trial justifies the additional spend.

GPT-6 Astra · 2026-09-30T17:37:41.000Z · View source GPT-5.5 · 2026-09-30T17:37:41.000Z · View source

Context window
1050000 · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Official feature documentation; no same-task quality measurement.
This is the documented context capacity, useful when testing a shared source packet across several documents. It does not guarantee retrieval accuracy, reliable citations, or live web access.
Cost
USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source
Standard ≤272K input tier; above that input ×2, output ×1.5. Caching/tools separate.
Use the recorded price as a budget input, not a total project quote. Usage, plan entitlements, taxes, and extra services can change the actual cost.

Strengths

Cost:
USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Standard ≤272K input tier; above that input ×2, output ×1.5. Caching/tools separate.

Trade-offs

Unreviewed dimensions remain open questions for your trial.

Cost: USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High
Standard ≤272K input tier; above that input ×2, output ×1.5. Caching/tools separate.
Plan-dependent. Check the official price and billing period.
Access: Local deployment:
Unknown
Official API:
Supported · Official fact · Current · View source
Workflow: Use the official API, then import the output into Proof Lab.
Commercial use and rights: Unknown
Rights depend on the product, plan, inputs and intended use. Review the linked terms.

Unknowns

Correctness · Source quality · Explanation quality

Evidence

Task support

Supported · Editorial assessment · Current · View source Last verified: 2026-09-30T17:29:41ZConfidence: Low

Context window

1050000 · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Official feature documentation; no same-task quality measurement.

Tool calling

Supported · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Official feature documentation; no same-task quality measurement.

Local deployment

UnknownLast verified: Not reviewedConfidence: Low

Official API

Supported · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Publicly documented channel; account, region, entitlement and quota were not tested.

Cost

USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-30T17:37:41.000ZConfidence: High

Standard ≤272K input tier; above that input ×2, output ×1.5. Caching/tools separate.

Commercial use

UnknownLast verified: Not reviewedConfidence: Low