DoubleChecks

What was claimed

New Anthropic Fable 5.1 model achieves 90% on ARC-AGI-2 and 97.5% on ARC-AGI-1 at much lower cost per task, representing major breakthrough in AGI progress

Our verdict

Needs caution

Sources support that Anthropic says Fable 5.1 reduces typical workload costs versus Fable 5, and one source lists ARC-AGI task costs, but the claim does not specify the baseline or exact comparison. The wording implies a verified large cost advantage that is not fully established by the provided evidence. This is a qualitative judgment, not a measurable fact. The sources show strong benchmark performance, but they do not establish an objective threshold for calling it a breakthrough in AGI. (Only 2 of 3 AI systems responded.)

Combined assessment29 sourcesChecked Sep 2, 2026

Why this verdict

  • It does so at much lower cost per task

    Misleading
    One assessment reached this conclusion; one differed·Claude: Verified

  • This represents a major breakthrough in AGI progress

    Can’t verify
    One assessment reached this conclusion; one differed·Claude: Misleading

  • New Anthropic Fable 5.1 model achieves 97.5% on ARC-AGI-1

    Verified
    One assessment reached this conclusion

  • New Anthropic Fable 5.1 model achieves 90% on ARC-AGI-2

    Verified
    All assessments reached this conclusion

Evidence reviewed

29 sources

Perplexity and Claude sources are combined here.

arcprize.organthropic.combenchlm.aivectorwire.aianotherwrapper.comanotherwrapper.com163.comaitooltier.comarcprize.orgx.com

Individual assessments

Assessment summary

  • The benchmark scores in the claim are supported by ARC Prize results for Anthropic Claude Fable 5.1.
  • The “much lower cost per task” part is directionally supported by Anthropic’s pricing notes, but the exact per-task cost comparison in the claim is not fully verified from the provided sources, and the “major breakthrough in AGI progress” language is subjective rather than a factual measurement.

Claims reviewed

It does so at much lower cost per task

Misleading
Sources support that Anthropic says Fable 5.1 reduces typical workload costs versus Fable 5, and one source lists ARC-AGI task costs, but the claim does not specify the baseline or exact comparison. The wording implies a verified large cost advantage that is not fully established by the provided evidence.

This claim may mislead readers without added context.

This represents a major breakthrough in AGI progress

Can’t verify
This is a qualitative judgment, not a measurable fact. The sources show strong benchmark performance, but they do not establish an objective threshold for calling it a breakthrough in AGI.

New Anthropic Fable 5.1 model achieves 90% on ARC-AGI-2

Verified
ARC Prize lists Anthropic Claude Fable 5.1 at 90.0% on ARC-AGI-2. The same result appears in benchmark aggregations based on Anthropic’s system card.

Assessment summary

  • The claim's core factual assertions are accurate.
  • Fable 5.1 does achieve the stated ARC-AGI scores at the specified costs, and the model does represent performance improvements.
  • The characterization as a "major breakthrough" is subjective but supported by Anthropic's own framing and industry coverage.

Claims reviewed

Represents major breakthrough in AGI progress

Misleading
While Fable 5.1 shows significant performance improvements on benchmarks, characterizing it as a 'major breakthrough in AGI progress' overstates the significance. The improvements are incremental refinements to an existing model class, not a fundamental breakthrough. Anthropic frames it as setting new standards on specific tasks, not as AGI progress.

This claim may mislead readers without added context.

Much lower cost per task

Verified
Fable 5.1 achieves these scores at $1.40 per ARC-AGI-1 task and $4.49 per ARC-AGI-2 task. At lower effort settings, Fable 5.1 delivers similar performance to Fable 5 at substantially lower cost (25% typical, up to 45% for agentic work). However, at max effort, cost per token actually increases due to higher output token usage.

Anthropic Fable 5.1 achieves 90% on ARC-AGI-2

Verified
Official ARC Prize leaderboard confirms Fable 5.1 scores 90.0% on ARC-AGI-2 Semi-Private at max effort.
ChatGPTDid not respond

Check your own claim

Paste any statement, headline, or AI answer. Independent assessments check it and compare sources.

Share this result