OpenGPTAI decision workspace
Search⌘K
Build software

Debug code

Refine candidates (optional)
Advanced requirements

Candidates to compare

Based on documented capabilities, not measured performance on your task. Display order is not a quality ranking.

What to test first

Use the same small change request and repository context. Run the tests, inspect the diff, and compare defects and the amount of manual repair required.

Claude Fable 5.1

Anthropic · platform.claude.com

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

  • Context window[1]

    1,000,000 tokens

    Supply code and related documentation in one prompt.

  • Cost[2]
    USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply

    Plan your trial budget using the recorded price.

Full evidence and sources

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 · 2026-09-09 · View source Claude Opus 5 · 2026-09-09 · View source
  • Context window
    1000000 · Official fact · Current · View source Last verified: 2026-09-09Confidence: High

    This capacity can accommodate code and related documentation in one prompt. Test whether it actually tracks dependencies and produces changes that pass your tests.

  • Cost
    USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source

    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: 2026-09-09Confidence: High

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: 2026-09-09Confidence: High
    Plan-dependent. Check the official price and billing period.
    Access
    Official API:
    Supported · Official fact · Current · View source
    Local deployment:
    Unknown
    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 · Maintainability · Explanation quality

    Evidence

    Tool calling
    Supported · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
    Context window
    1000000 · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
    Task support
    Supported · Editorial assessment · Current · View source Last verified: 2026-09-09Confidence: Low
    Official API
    Supported · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
    Local deployment
    UnknownLast verified: Not reviewedConfidence: Low
    Cost
    USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
    Commercial use
    UnknownLast verified: Not reviewedConfidence: Low

    Claude Opus 5

    Anthropic · platform.claude.com

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

    • Context window[1]

      1,000,000 tokens

      Supply code and related documentation in one prompt.

    • Cost[2]
      USD 5 input / 25 output per 1M tokens; standard tier, additional charges may apply

      Plan your trial budget using the recorded price.

    Full evidence and sources

    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 · 2026-09-09 · View source Claude Fable 5.1 · 2026-09-09 · View source
    • Context window
      1000000 · Official fact · Current · View source Last verified: 2026-09-09Confidence: High

      This capacity can accommodate code and related documentation in one prompt. Test whether it actually tracks dependencies and produces changes that pass your tests.

    • Cost
      USD 5 input / 25 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source

      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: 2026-09-09Confidence: High

    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: 2026-09-09Confidence: High
      Plan-dependent. Check the official price and billing period.
      Access
      Official API:
      Supported · Official fact · Current · View source
      Local deployment:
      Unknown
      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 · Maintainability · Explanation quality

      Evidence

      Tool calling
      Supported · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
      Context window
      1000000 · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
      Task support
      Supported · Editorial assessment · Current · View source Last verified: 2026-09-09Confidence: Low
      Official API
      Supported · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
      Local deployment
      UnknownLast verified: Not reviewedConfidence: Low
      Cost
      USD 5 input / 25 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
      Commercial use
      UnknownLast verified: Not reviewedConfidence: Low

      Gemini 3.7 Flash

      Google · ai.google.dev

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

      • Context window[1]

        1,048,576 tokens

        Supply code and related documentation in one prompt.

      • Cost[2]
        USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply

        Plan your trial budget using the recorded price.

      Full evidence and sources

      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 · 2026-09-09 · View source Gemini 3.8 Flash · 2026-09-09 · View source
      • Context window
        1048576 · Official fact · Current · View source Last verified: 2026-09-09Confidence: High

        This capacity can accommodate code and related documentation in one prompt. Test whether it actually tracks dependencies and produces changes that pass your tests.

      • Cost
        USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source

        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: 2026-09-09Confidence: High

      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: 2026-09-09Confidence: High
        Plan-dependent. Check the official price and billing period.
        Access
        Official API:
        Supported · Official fact · Current · View source
        Local deployment:
        Unknown
        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 · Maintainability · Explanation quality

        Evidence

        Tool calling
        Supported · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
        Context window
        1048576 · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
        Task support
        Supported · Editorial assessment · Current · View source Last verified: 2026-09-09Confidence: Low
        Official API
        Supported · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
        Local deployment
        UnknownLast verified: Not reviewedConfidence: Low
        Cost
        USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
        Commercial use
        UnknownLast verified: Not reviewedConfidence: Low

        Gemini 3.8 Flash

        Google · ai.google.dev

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

        • Context window[1]

          1,048,576 tokens

          Supply code and related documentation in one prompt.

        • Cost[2]
          USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply

          Plan your trial budget using the recorded price.

        Full evidence and sources

        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 · 2026-09-09 · View source Gemini 3.7 Flash · 2026-09-09 · View source
        • Context window
          1048576 · Official fact · Current · View source Last verified: 2026-09-09Confidence: High

          This capacity can accommodate code and related documentation in one prompt. Test whether it actually tracks dependencies and produces changes that pass your tests.

        • Cost
          USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source

          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: 2026-09-09Confidence: High

        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: 2026-09-09Confidence: High
          Plan-dependent. Check the official price and billing period.
          Access
          Official API:
          Supported · Official fact · Current · View source
          Local deployment:
          Unknown
          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 · Maintainability · Explanation quality

          Evidence

          Tool calling
          Supported · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
          Context window
          1048576 · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
          Task support
          Supported · Editorial assessment · Current · View source Last verified: 2026-09-09Confidence: Low
          Official API
          Supported · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
          Local deployment
          UnknownLast verified: Not reviewedConfidence: Low
          Cost
          USD 0.75 input / 3.75 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
          Commercial use
          UnknownLast verified: Not reviewedConfidence: Low

          GPT-5.5

          OpenAI · developers.openai.com

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

          • Context window[1]

            1,050,000 tokens

            Supply code and related documentation in one prompt.

          • Cost[2]
            USD 5 input / 30 output per 1M tokens; standard tier, additional charges may apply

            Plan your trial budget using the recorded price.

          Full evidence and sources

          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 · 2026-09-09 · View source GPT-6 Astra · 2026-09-09 · View source
          • Context window
            1050000 · Official fact · Current · View source Last verified: 2026-09-09Confidence: High

            This capacity can accommodate code and related documentation in one prompt. Test whether it actually tracks dependencies and produces changes that pass your tests.

          • Cost
            USD 5 input / 30 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source

            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: 2026-09-09Confidence: High

          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: 2026-09-09Confidence: High
            Plan-dependent. Check the official price and billing period.
            Access
            Official API:
            Supported · Official fact · Current · View source
            Local deployment:
            Unknown
            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 · Maintainability · Explanation quality

            Evidence

            Tool calling
            Supported · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
            Context window
            1050000 · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
            Task support
            Supported · Editorial assessment · Current · View source Last verified: 2026-09-09Confidence: Low
            Official API
            Supported · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
            Local deployment
            UnknownLast verified: Not reviewedConfidence: Low
            Cost
            USD 5 input / 30 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
            Commercial use
            UnknownLast verified: Not reviewedConfidence: Low

            GPT-6 Astra

            OpenAI · developers.openai.com

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

            • Context window[1]

              1,050,000 tokens

              Supply code and related documentation in one prompt.

            • Cost[2]
              USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply

              Plan your trial budget using the recorded price.

            Full evidence and sources

            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 · 2026-09-09 · View source GPT-5.5 · 2026-09-09 · View source
            • Context window
              1050000 · Official fact · Current · View source Last verified: 2026-09-09Confidence: High

              This capacity can accommodate code and related documentation in one prompt. Test whether it actually tracks dependencies and produces changes that pass your tests.

            • Cost
              USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source

              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: 2026-09-09Confidence: High

            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: 2026-09-09Confidence: High
              Plan-dependent. Check the official price and billing period.
              Access
              Official API:
              Supported · Official fact · Current · View source
              Local deployment:
              Unknown
              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 · Maintainability · Explanation quality

              Evidence

              Tool calling
              Supported · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
              Context window
              1050000 · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
              Task support
              Supported · Editorial assessment · Current · View source Last verified: 2026-09-09Confidence: Low
              Official API
              Supported · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
              Local deployment
              UnknownLast verified: Not reviewedConfidence: Low
              Cost
              USD 10 input / 50 output per 1M tokens; standard tier, additional charges may apply · Official fact · Current · View source Last verified: 2026-09-09Confidence: High
              Commercial use
              UnknownLast verified: Not reviewedConfidence: Low