|

Anthropic Releases Claude Fable 5.1 and Claude Mythos 5.1: 52.6% on Terminal-Bench-Science and 75% Cheaper Cache Reads

ByRicardo

Anthropic has launched Claude Fable 5.1 and Claude Mythos 5.1, three months after the Fable 5 line shipped in June 2026. The two are the identical underlying mannequin behind completely different safeguard layers. Fable 5.1 is usually accessible as claude-fable-5-1; Mythos 5.1 stays restricted to vetted organizations. Both carry a 1M token context window and 128K max output tokens, with adaptive pondering all the time on. The headline functionality quantity is 52.6% on Terminal-Bench-Science 0.1, towards 24.7% for Fable 5 and 29.0% for Opus 5. The headline industrial quantity is a 75% reduce to cache reads, from $1.00 to $0.25 per million tokens, which Anthropic measures as roughly 25% decrease value on typical workloads and as much as 45% on agentic ones. Base enter and output pricing is unchanged at $10 and $50 per million.

Is it deployable?

Yes, Claude Fable 5.1 is usually accessible as claude-fable-5-1 on the Claude API, Amazon Bedrock, Claude Platform on AWS, Google Cloud, and Microsoft Foundry. Claude Mythos 5.1 is just not: it’s restricted to vetted US organizations inside Project Glasswing.

The benchmark numbers

On Terminal-Bench-Science 0.1, an agentic scientific analysis benchmark, Fable 5.1 scores 52.6% towards 29.0% for Opus 5, 24.7% for Fable 5, and 22.4% for GPT-5.6 Sol. Anthropic reviews a regular error of three.5 to 4.5 factors per mannequin, so deal with the margin, not the rating, with care.

On Terminal-Bench 4.0, Fable 5.1 reaches 55.8% and Mythos 5.1 reaches 60.9%. The hole between two an identical fashions is the price of safeguard interventions, which is an unusually trustworthy disclosure. Elsewhere: CursorBench 3.2.0 at 73.4%, Humanity’s Last Exam at 60.9% with out instruments and 65.0% with instruments, AutomationBench at 31.4%, OSWorld 2.0 at 41.7% strict, and GDPval-AA v2 at 1853.

Where the price reduce comes from

Base enter and output pricing is unchanged. Cache reads drop 75%, from $1.00 to $0.25 per million tokens, which is 0.025 instances base enter towards 0.1 on each different Claude mannequin. Anthropic measures roughly 25% decrease value on typical workloads and as much as about 45% on context-heavy agentic ones. Batch processing is $5 and $25 per million tokens.

Three breaking modifications groups will hit

  1. Forced device use is gone: tool_choice set to any or device returns a 400. Use auto with strict device use or structured outputs as a substitute.
  2. Thinking blocks are model-bound: Fable 5.1 reads earlier fashions’ pondering, however no earlier mannequin reads its personal. Router and fallback setups lose reasoning once they change down.
  3. Editing earlier turns invalidates pondering blocks: Injecting and deleting per-turn reminders, or rebuilding the system or instruments array mid-conversation, now errors. The verify is enforced for accounts created on or after August 31, 2026. The fixes are turn-scoped system messages and server-side context enhancing.

Additive modifications: per-message effort, turn-scoped system messages, and pondering.show: "updates" are all in beta behind headers. Content provenance is just not elective, with a statistical textual content watermark on all output and C2PA credentials on information.

Anthropic additionally paperwork actual regressions. Parallel device calling is extra variable, so agent loops might subject one name per flip the place Fable 5 batched a number of. The mannequin narrates much less, solutions from reminiscence extra typically at low effort, and prefers whole-file rewrites over focused edits.

Safeguards and science

Cyber safeguards now allow vulnerability discovery however not exploit improvement, reducing interventions in Claude Code by roughly 60% per session. Biology safeguards fireplace 85% much less typically on benign requests. Penetration testing, exploit era, and binary-based vulnerability scanning nonetheless redirect to Opus.

On analysis, Mythos 5.1 designed protein binders with roughly 50% hit charge throughout 12 targets towards a ten to fifteen% norm, Fable 5.1 constructed a Venus elevation map at 2 to three km decision, and customized GPU kernels sped up seven open-source genomics fashions by as much as 2.5x.


Key Takeaways

  • Fable 5.1 and Mythos 5.1 are one mannequin behind two safeguard layers; solely Fable is usually accessible.
  • 52.6% on Terminal-Bench-Science 0.1, roughly double Fable 5’s 24.7%.
  • Cache reads fall 75% to $0.25 per million; base charges are untouched.
  • Three breaking API modifications hit any agent that edits dialog historical past.
  • 30-day retention applies; zero information retention wants categorical authorization.

Check out the Technical Details. Also, be at liberty to observe us on Twitter and don’t neglect to affix our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.

Need to companion with us for selling your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar and so on.? Connect with us

The publish Anthropic Releases Claude Fable 5.1 and Claude Mythos 5.1: 52.6% on Terminal-Bench-Science and 75% Cheaper Cache Reads appeared first on MarkTechPost.

Similar Posts