Fable 5.1 Launches With Massive Benchmark Jumps Despite Minor Version Bump
Fable Studio (Anthropic) released Fable 5.1 on 09-02. Despite the mere 0.1 version bump, multiple evaluators reported a step-change improvement across several benchmarks plus a 25% cost reduction, making it a far stronger release than the version number suggests.
Confirmed
- Scores 52.6% on Terminal-Bench-Science 0.1, doubling Fable 5's 24.7%, with gains concentrated in agentic scientific research (@markk, @DrSingularity)
- Scores 55.8% on Terminal-Bench 4.0, beating Fable 5's 42% and Opus 5's 52.3%, and also leading in agentic coding and computer-use evaluations (@firstadopter)
- Costs 25% less than the previous generation (@markk)
- @cedricchee noted Fable 5.1's progress on the Artificial Analysis index is bigger than it appears
Why it matters
- @DrSingularity considers "agentic scientific research" a key benchmark on the path to 1000x faster, AI-driven scientific progress, and the doubling is seen as a major milestone in that direction
- Both @kimmonismus and @firstadopter see this as competitive pressure that OpenAI and Astra must answer head-on
2026-09-02 ~ 2026-09-02 · 7 related posts
- Episode 1: Claude Fable 5.1 Spotted on Bedrock and in Docs Ahead of Imminent Launch(2026-09-01, 5 posts)
- Episode 2: Anthropic Launches Claude Fable 5.1 with 75% Cheaper Caching and Secure Mythos 5.1(2026-09-02, 35 posts)
- Episode 3: Fable 5.1 Launches With Massive Benchmark Jumps Despite Minor Version Bump(2026-09-02, 7 posts)
- Episode 4: Every's Hands-On: Anthropic's Fable 5.1 Is the Best Coding Model at Half the Token Cost(2026-09-02, 7 posts)
- Episode 5: Claude Fable 5.1 Low-Effort Mode Matches Opus at Lower Cost(2026-09-02, 2 posts)
Primary sources
- Fable 5.1 release: Doubles scientific agent scores, cuts costs by 25% — mark_k ·
- Fable 5.1 Released: Amazing Progress in Agentic Scientific Research — Dr_Singularity ·
- Fable 5.1 beats Opus 5 in coding benchmarks — firstadopter ·
- Fable 5.1 benchmarks show insane jumps on Terminal/Science — kimmonismus · 2026-09-02
- [source] Fable 5.1 release: Doubles scientific agent scores, cuts costs by 25% — mark_k · 2026-09-02
- [source] Fable 5.1 Released: Amazing Progress in Agentic Scientific Research — Dr_Singularity · 2026-09-02
- Fable 5.1 scores 52.6% on Terminal-Bench-Science, doubling previous performance — Dr_Singularity · 2026-09-02
- [source] Fable 5.1 beats Opus 5 in coding benchmarks — firstadopter · 2026-09-02
- Fable 5.1 makes a bigger jump than it looks on the Artificial Analysis Index — cedric_chee · 2026-09-02
- Fable 5.1's real gains are bigger than its Artificial Analysis index jump suggests — cedric_chee · 2026-09-02