Claude Fable 5.1 is here — announced today, September 1, 2026 — and it ships with a price cut that changes the economics of frontier coding. Same $10 / $50 per million tokens as Fable 5 on the surface, but cache reads drop 75% to $0.25 per million tokens, which Anthropic says cuts the cost of typical workloads by an estimated 25% and highly agentic workloads by up to roughly 45%.
Alongside it, Anthropic released Claude Mythos 5.1 — the same underlying model with stricter safeguards, reserved for its trusted access programs in cybersecurity and the life sciences. This is Anthropic's fifth-generation flagship refresh: more capable than Fable 5 across coding, knowledge work, and long-running problem-solving, with benchmark numbers that pulled far ahead of the predecessor on agentic tasks while the total bill for running it went down.
Pricing: The Headline Is a Cache Read
Fable 5.1 is priced at $10 per million input tokens and $50 per million output tokens — unchanged from Fable 5. The entire price cut lives in one line item:
| Pricing component | Fable 5 | Fable 5.1 | Change |
|---|---|---|---|
| Input tokens | $10 / 1M | $10 / 1M | Unchanged |
| Output tokens | $50 / 1M | $50 / 1M | Unchanged |
| Cache reads | $1.00 / 1M | $0.25 / 1M | −75% |
| Estimated typical workload cost | — | −25% vs. Fable 5 | Cheaper |
| Highly agentic workload cost | — | up to −45% vs. Fable 5 | Much cheaper |
Cache reads are the mechanism: Fable 5.1 reads inputs that have already been processed and stored at a quarter of the previous price. For agent loops that repeatedly replay a large system prompt, codebase context, or tool definitions, prompt caching is not a nicety — it is most of the bill. Cutting that price is precisely aimed at the workloads the model was built for. Anthropic's own framing is that at Low or Medium effort, Fable 5.1 achieves similar or better results than Fable 5 at a much lower cost; Fable 5.1 defaults to High effort in Claude Code and Medium in Claude Cowork and on Claude.ai.
US-only inference is available at 1.1x pricing for input and output tokens for workloads that must run in the US.
Benchmarks: The Gap Is at the Agentic Edge
The interesting numbers are not where Fable 5.1 is merely best — they are where it is far ahead. On Terminal-Bench-Science 0.1 it more than doubled Fable 5's score, and on AutomationBench it nearly doubled it, which is a strong signal about what actually changed in this release: long-horizon, tool-using, scientific and business workflows.
| Benchmark | Fable 5.1 | Fable 5 | Opus 5 | GPT-5.6 Sol |
|---|---|---|---|---|
| Terminal-Bench-Science 0.1 | 52.6% | 24.7% | 29.0% | 22.4% |
| Terminal-Bench 4.0 | 55.8% (60.9% Mythos) | 42.0% | 52.3% | 37.3% |
| GDPval-AA v2 (knowledge work) | 1853 | 1723 | 1824 | 1711 |
| CursorBench 3.2.0 | 73.4% | 70.5% | 70.0% | 67.2% |
| OSWorld 2.0 (partial) | 77.9% | 72.9% | 75.4% | — |
| OSWorld 2.0 (strict) | 41.7% | 36.1% | 39.6% | — |
| Humanity's Last Exam (no tools) | 60.9% | 57.8% | 56.6% | — |
| Humanity's Last Exam (with tools) | 65.0% | 63.8% | 63.6% | — |
| AutomationBench | 31.4% | 17.1% | 26.9% | 19.6% |
Every row in the table is Anthropic's own published evaluation, and the caveat that should be attached to any vendor benchmark applies — treat these as Anthropic's claims, not independent reproduction. But within that frame, the pattern is consistent: Fable 5.1 leads every row, and the largest margins appear exactly where agentic and long-horizon work lives rather than in single-shot Q&A.
The Same Model Behind Two Doors
Fable 5.1 and Mythos 5.1 are the same model with different levels of safeguards. Fable 5.1 is generally available to Pro, Max, Team, and Enterprise users, and to developers through the Claude Platform natively, marketplaces, AWS, Google Cloud, and Microsoft Foundry. Mythos 5.1 is available only through trusted access programs, with safeguards designed for cybersecurity and the life sciences.
The safety story got materially better with this release: in cybersecurity, the newest safeguards block 60% fewer false positives than before — meaning fewer benign queries get bounced to weaker models. Notably, Fable 5.1 can now discover software vulnerabilities, though it cannot develop exploits for them — a capability that opens the door for safety research while keeping exploit development out of reach.
In biology, Anthropic established an access program developed in partnership with the US government to enable access to Mythos 5.1's advanced biology capabilities, with enrollment for scientists expected to open soon.
What Customers Saw
Anthropic shared early-customer feedback, and the recurring themes are exactly the ones the pricing points at: long-horizon reliability and lower cost per task.
- Millennium (investment firm) tested Fable 5.1 in internal systems and reported it found the cause of a rare crash — roughly one-in-a-million-runs — that none of their engineers (or any other model) had explained in four to five years. Fable 5.1 disassembled an external vendor library, matched it against the core dump, and traced the crash to a bug in that library.
- Devin said it is moving its Opus 5 traffic to Fable 5.1 on launch day — it matched or edged out Fable 5 at a lower cost per task, and with the new cache-read pricing a Fable-class model finally becomes economical for workloads they'd kept on Opus, starting with code review.
- One customer reported Fable 5.1 is about twice as fast as Opus 5 and uses half as many tokens, and another described it building a complex prototype over about three days — research across services code and docs, then hours of unattended work with verification loops, delivering a full visual walkthrough.
And in science, Anthropic shared early evidence of genuine research contribution: Mythos 5.1 designed high-affinity protein binders with affinities roughly 10 times higher than the best Adaptyv Bio competition designs (hit rate near 50% across 12 targets versus a typical 10–15%), and produced a new high-resolution elevation map of a third of Venus from 30-year-old Magellan radar data — resolving features down to 2–3km instead of 10–20km and heights up to 25% more accurate.
Data Retention and Enterprise Frontier Safeguards
Fable 5.1 defaults to 30-day data retention for safety monitoring. The bigger news is a new system — Enterprise Frontier Safeguards (EFS) — that stores data in cloud infrastructure controlled entirely by the customer rather than Anthropic, delivering complete privacy equivalent to a zero-data-retention policy while remaining state-of-the-art at preventing adversarial use. EFS will roll out to enterprise customers in phases beginning later this fall; until it is available, eligible customers can use Fable 5.1 with zero data retention.
The Takeaway
Fable 5.1 is the rare launch where the most capable model is also the cheaper one to run. The $10/$50 rate card looks flat, but the 75% cut to cache reads re-prices the workloads that actually consume tokens — agent loops, long context, repeated tool calls — by 25% typical and up to 45% at the agentic tail. The benchmark table backs the positioning: better everywhere, and dramatically better precisely on the long-horizon, tool-heavy tasks that define 2026-style agent work. If you are paying per token for agentic coding, this is the first Fable-class model where the economics stop fighting you.
Sources: Anthropic announcement — Claude Fable 5.1 & Mythos 5.1 · Anthropic Claude Fable · Anthropic Fable 5.1 system card · 9to5Mac · OpenRouter
Comments
Leave a comment
Related articles
Claude Max Weekly Limits: The 50% Summer Boost Becomes a Permanent 25% Raise — Math Included
Anthropic's temporary 50% Claude Code weekly boost ends September 13. On September 14 a permanent 25% increase to baseline limits takes over — a net ~17% cut for active users. The math, the timeline, what stays doubled.
Ox Alpha: The Anonymous 1M-Context Coding Model That Has Everyone Guessing Z.ai
Ox Alpha is a free, anonymous 1M-context reasoning model on OpenRouter. Community fingerprinting points to Zhipu AI's GLM-5.x. We break down the benchmarks, the mystery, and the adoption.
DeepSeek V4 Pro Is Now an Ollama Cloud Model: The 1.6T Flagship on a $20 Flat Plan
DeepSeek's 1.6T flagship V4 Pro is now a flat-rate Ollama Cloud model — $20 a month, one line of shell, drop-in for Claude Code and OpenCode. Here's what the 0813 refresh changed and when the plan beats the API.