Claude Fable 5.2 vs GPT-⁠6 Astra.

Claude Fable 5.2 scores higher, 65 against 57, but GPT-⁠6 Astra is more than 4 times cheaper per task.

Anthropic and OpenAI
Data from 13 November 2026.

Which is better?

Claude Fable 5.2 scores higher on Artificial Analysis's Intelligence Index, 65 against 57. It is 1st of the 58 models it scores, and GPT-⁠6 Astra 4th.

On the four coding and reasoning tests, Claude Fable 5.2 wins FrontierCode 2.0, HLE Diamond and CritPt v2, and they are level on Terminal-⁠Bench 4.0.

Per task, GPT-⁠6 Astra is more than 4 times cheaper and faster too, $6.24 against $26.54 and 24.2 minutes against 46.8 minutes.

Per token, Anthropic charges $8 per million input tokens and $40 per million output tokens, and OpenAI $10 and $50. Claude Fable 5.2 is cheaper per token, but it costs more for each task in the index. They read about the same amount at once, 1 million and 1.05 million tokens.

  • Choose Claude Fable 5.2 if

    • you want the higher score, 65 against 57
    • you want it for reasoning, it wins HLE Diamond and CritPt v2
    • you want it to change code in real projects, it scores 48.8% against 42.4% on FrontierCode 2.0
  • Choose GPT-⁠6 Astra if

    • the cost of each task matters, $6.24 against $26.54
    • you want it to take less time, 24.2 minutes against 46.8 per task

At a glance

FigureClaude Fable 5.2AnthropicGPT-⁠6 AstraOpenAI
ScoreArtificial Analysis's index65Better1st of 58574th of 58
Cost per task$26.5451st of 51$6.24Better44th of 51
Time per task46.8 min46th of 5024.2 minBetter33rd of 50
Input pricePer million tokens$8Better$10
Output pricePer million tokens$40Better$50
Context1 million1.05 million
WeightsClosedClosed
Effort measuredEach one's highestMaxMax
Released9 Nov 20263 Sept 2026

The score, the cost and the time per task and the tests from Artificial Analysis, read on 13 November 2026 with each model at its highest effort. The price, the context and the release date from Anthropic's and OpenAI's pages. The table of the 58.

Their tests

Coding

FrontierCode 2.0 Changes to real open-⁠source projects.

  • Claude Fable 5.248.8%
  • GPT-⁠6 Astra42.4%

Terminal-⁠Bench 4.0 Complex work in the terminal, on its own.

  • Claude Fable 5.259.6%
  • GPT-⁠6 Astra59.1%

Reasoning

HLE Diamond Expert-⁠level questions.

  • Claude Fable 5.280.2%
  • GPT-⁠6 Astra63.8%

CritPt v2, hard subset Physics research problems.

  • Claude Fable 5.276.5%
  • GPT-⁠6 Astra70.5%

Questions.

  • Which one is cheaper?

    Per task, GPT-⁠6 Astra, with $6.24 against $26.54 on the tasks of Artificial Analysis's index. Per token, Claude Fable 5.2, with $8 and $40 per million input and output tokens against $10 and $50. The cost per task says more than the price per million tokens, because for the same work some models use many more tokens than others.

  • Which one reads more text at once?

    They read about the same amount at once, 1 million and 1.05 million tokens.

  • Can they run on your own servers?

    No. Neither has its weights published, so they are used through Anthropic and OpenAI or whoever offers them.

  • Which one for your company's software?

    It depends on the job. We cover it in “Which AI Model to Use to Build Your Company's Software”.

Let's talk.

Tell us what you need and one of the partners who write the code gets back to you.