Anthropic Launches Claude Fable 5.1: Big Benchmark Gains and 75% Cache Price Cut, but Real-World Costs Divide Users

On September 2, Anthropic officially released Claude Fable 5.1, positioned as the latest upgrade to its most powerful model family, specialized in complex, long-running tasks and research scenarios. Alongside it, the company launched Claude Mythos 5.1, available only through a trusted-access program (for cyber defense and life sciences). Both share the same underlying model, differing only in safety guardrail levels. Anthropic also announced a reset of all users' 5-hour and weekly usage limits, and disclosed in support docs changes to thinking-block management in the Messages API to prevent model distillation.

Confirmed

• Pricing: Input/output stays at $10/$50 per million tokens; Prompt Cache read costs drop 75% (to $0.25); Anthropic estimates typical workloads will see roughly 25% lower costs, with even bigger savings in cache-heavy scenarios
• Benchmarks: Terminal-Bench-Science 0.1 score of 52.6% (double that of gen 5); kimmonismus relayed that Terminal-Bench 4.0, Science-Bench, HLE, and other benchmarks all jumped significantly; results posted by Anthropic engineer Felix Rieseberg show incremental progress over the previous generation, while Justwalkingthru3 views the gains as limited and predicts OpenAI's upcoming Astra could soon overtake it
• Value: kimmonismus noted that on Cursor Bench, Fable 5.1 High clearly beats Sol 5.6 Max while costing less; Anthropic says lower effort tiers can surpass Fable 5 at lower cost
• Conflicting cost reports: MineBench maintainer ENTAlam found API pricing unchanged, but total cost across 15 builds rose from $54.93 to $147.55 (nearly 3x) with doubled inference time; WonderFactory also shared a screenshot showing a single task costing $3.69, higher than Fable 5; meanwhile, developer thispageleftblank, based on their own data (cache reads = 78% of spend), estimated total costs could drop about 57%
• User feedback: Reddit users (relayed by KrayeBaby) say multi-turn session output is more coherent and readable than Opus 5 and Fable 5, while acknowledging the model still has flaws

Why it matters

Fable 5.1 is Anthropic's positioning move just ahead of OpenAI's new model (Astra) — Wes Roth also noted the release timing closely follows OpenAI's moves. The steep cache price cut reshapes costs for heavy users, but real-world tests like MineBench suggest the model may burn more tokens through longer thinking, so whether you actually save money depends heavily on the task profile — the key variable in judging this round of price cuts.

2026-09-02 ~ 2026-09-03 · 166 related posts

Full story(3 episodes)→

• Episode 1: Claude Fable 5.1 Surfaces on Bedrock and in Official Docs, Launch Imminent(2026-09-01, 5 posts)
• Episode 2: Anthropic Launches Claude Fable 5.1: Big Benchmark Gains and 75% Cache Price Cut, but Real-World Costs Divide Users(2026-09-02, 166 posts)
• Episode 3: Fable 5.1 Users Report Surging Quota Usage and Suspected Claude Code Cache Bug(2026-09-02, 6 posts)

• Anthropic releases Fable 5.1, updates API thought block handling — Icecream_monday · 2026-09-02
• Fable 5.1 model released with unchanged pricing — scaling01 · 2026-09-02
• Fable 5.1 now available in Claude API — Angaisb_ · 2026-09-02
• Fable 5.1 benchmarks released — scaling01 · 2026-09-02
• Claude Fable 5.1 Launches with Better Benchmarks and 75% Cheaper Cache — Anthropic · 2026-09-02
• Anthropic: Fable 5.1 excels at long-running tasks and early scientific research — claudeai · 2026-09-02
• Anthropic releases Claude 5.1: 25% cheaper with zero-retention enterprise safeguards — claudeai · 2026-09-02
• Claude Fable 5.1 Launches: High Usage Hints at Major Compute Jump — VraserX · 2026-09-02
• Anthropic Launches Claude Fable 5.1 and Mythos 5.1 Models — AI_Andrew · 2026-09-02
• [source] Anthropic releases Fable 5.1 and Mythos 5.1 with major cost cuts — ClaudeOfficial · 2026-09-02
• Anthropic Fable 5.1 is now available in Claude Code v2.1 — BLUECOW009 · 2026-09-02
• Fable 5.1 benchmarks show insane jumps on Terminal/Science — kimmonismus · 2026-09-02
• Every's Vibe Check: Anthropic's Fable 5.1 Reclaims the Coding Crown — every · 2026-09-02
• Anthropic Is Back? Fable 5.1 Hands-On: Major Gains in Coding and Writing — every · 2026-09-02
• Anthropic releases Claude 5.1 for coding and knowledge work — kieranklaassen · 2026-09-02
• Claude Fable 5.1 Scores Leak: 77.9% on OSWorld, Beating GPT-5.6 Sol Across Benchmarks — Scobleizer · 2026-09-02
• Benchmarks for Claude Fable 5.1 and Claude Mythos 5.1 leaked — minxio_ · 2026-09-02
• Anthropic quietly launches Claude Fable 5.1 for "hardest challenges" — External-Grape-2978 · 2026-09-02
• Claude new model benchmark leak: 52.6% on TBS-Science — eyishazyer · 2026-09-02
• Fable 5.1 release: Doubles scientific agent scores, cuts costs by 25% — mark_k · 2026-09-02
• Fable 5.1 improves cybersecurity capabilities with significant reduction in fallbacks — ClaudeDevs · 2026-09-02
• Fable 5.1 Released: Amazing Progress in Agentic Scientific Research — Dr_Singularity · 2026-09-02
• Mazebench Launches to Stress-Test Fable 5.1 with Billions of Tokens — patience_cave · 2026-09-02
• Claude Code 2.1.257: Defaults to Fable 5.1, Hardens Auto Mode Security — ClaudeCodeLog · 2026-09-02
• Claude Code 2.1.257 Released: Defaults to Fable 5.1 — ClaudeCodeLog · 2026-09-02
• Claude Fable 5.1 Doubles Scores on Science Benchmarks — testingcatalog · 2026-09-02
• Claude Fable 5.1: Costs Drop 45%, Agent Benchmarks Double — kimmonismus · 2026-09-02
• Anthropic releases official introduction video for Claude Fable 5.1 — avilacjf · 2026-09-02
• Anthropic Launches Claude Fable 5.1 with Better Coding and Cheaper Cache — Scobleizer · 2026-09-02
• Anthropic's Fable 5.1 Leaks: 2x Faster, Half Cost — Yuchenj_UW · 2026-09-02
• Fable 5.1 scores 52.6% on Terminal-Bench-Science, doubling previous performance — Dr_Singularity · 2026-09-02
• Fable 5.1 beats Opus 5 in coding benchmarks — firstadopter · 2026-09-02
• Dev Calc: Fable 5.1 Price Cut Could Reduce Total Costs by 57% — _thispageleftblank · 2026-09-02
• Fable 5.1 Pricing Details: 75% Cheaper Cache — ThunderStorm420 · 2026-09-02
• FrontierFinance eval: Fable 5.1 leads with 55.9% score, 1.7x cost increase — maithra_raghu · 2026-09-02
• Fable 5.1 Review: More Powerful, Faster, and Half the Cost — danshipper · 2026-09-02
• Tips for Claude Fable 5.1: Low-Effort Mode and Cost Optimization — RLanceMartin · 2026-09-02
• User Feedback: Fable 5.1 Outputs are Clearer, No More Neuralese Loops — Justin_Halford_ · 2026-09-02
• Claude Fable 5.1 Launches: Doubles Science Benchmarks, Solves Data Retention — dotey · 2026-09-02
• Fable 5.1 Launches: Doubled Performance at Half the Token Cost — danshipper · 2026-09-02
• Anthropic Releases Fable 5.1 with Massive Gains — AGI Hunt · 2026-09-02
• Fable 5.1 released: Agentic coding tasks solved at half the cost — airesearch12 · 2026-09-02
• Anthropic Releases Claude Fable 5.1 and Resets Usage Limits — thesaraharminta · 2026-09-02
• Anthropic Releases Claude Fable with Doubled Terminal-Bench Scores — BenBlaiszik · 2026-09-02
• Fable 5.1 launches with 75% cheaper cache reads — EricBuess · 2026-09-02

5 near-duplicate retellings: scaling01 · TFenrir · Scobleizer · prasenx · aziz4ai