Last updated: November 12, 2026 · Prices, limits and availability checked against Anthropic’s launch post, API docs, pricing page and Help Center on this date
Claude Fable 5.2 is Anthropic’s most capable generally available model, released on November 9, 2026 as claude-fable-5-2. It costs $8 per million input tokens and $40 per million output tokens, 20% less than Fable 5.1, with a 1M-token context window and 128K output. Anthropic says it is up to 30% faster than 5.1. Claude Mythos 5.2 is the same model without cyber and biology safeguards, for verified organizations only.
Anthropic shipped the third Fable model in five months on November 9. Fable 5.2 replaces Fable 5.1 at the top of the generally available lineup. It is cheaper per token, faster, and better on almost every benchmark Anthropic published. It also brings four API changes that can break code written for 5.1. This guide covers what changed, what it costs on the API and on Claude plans, which benchmark numbers are worth trusting, who can get Mythos 5.2, and how to migrate.
Every price and limit below comes from Anthropic’s own pages, checked on November 12, 2026. Benchmark figures are Anthropic’s, not independent tests, and we label them that way. For where Fable sits among the other current models, see our Claude models overview.
What is new in Claude Fable 5.2 compared with Fable 5.1?
According to the launch post and the “What’s new in Claude Fable 5.2” docs page, the main changes are:
- Lower price. $8 / $40 per million tokens, down from $10 / $50. Cache reads drop from $0.25 to $0.20.
- Faster output. Anthropic says Fable 5.2 and Mythos 5.2 produce output up to 30% faster than the 5.1 models.
- Better coding, reasoning and vision. The largest jumps Anthropic reports are in agentic coding, scientific terminal tasks, remote freelance-style work and chart reading (table below).
- Fewer false refusals. Anthropic reports about 55% fewer safety-classifier blocks than Fable 5.1: about 60% fewer in cybersecurity and 40% fewer in biology.
- Medium effort by default in the apps. Claude apps and Claude Code now run Fable at medium effort. Anthropic says 5.2 at medium finishes hard tasks with fewer tokens than 5.1, but publishes no exact figure. On the API the default stays
high. - Fast mode. A new optional mode with up to 2.5x higher output speed, at $16 input and $80 output per million tokens. Fable 5.1 had no fast mode.
What did not change: the 1M-token context window, 128K max output, always-on adaptive thinking, the tokenizer (the one introduced with Opus 4.7), and the 30-day data retention requirement.
How much does Claude Fable 5.2 cost per million tokens?
All figures are USD list prices from the Claude API pricing page on November 12, 2026. Mythos 5.2 has exactly the same rate card as Fable 5.2.
| Price per MTok | Fable 5.2 | Fable 5.1 | Opus 5.5 |
|---|---|---|---|
| Base input | $8.00 | $10.00 | $4.00 |
| Output | $40.00 | $50.00 | $20.00 |
| Cache write, 5 minutes | $10.00 | $12.50 | $5.00 |
| Cache write, 1 hour | $16.00 | $20.00 | $8.00 |
| Cache read (hit) | $0.20 | $0.25 | $0.20 |
| Batch API, input / output | $4 / $20 | $5 / $25 | $2 / $10 |
| Fast mode, input / output | $16 / $80 | not offered | $8 / $40 |
Fast mode can’t be combined with the Batch API. Prompt-caching and data-residency multipliers apply on top of fast-mode prices.
Three details matter for budgeting:
- The gap to Opus 5.5 shrank from 2.5x to 2x. Fable 5.1 cost 2.5 times Opus 5.5 per token. Fable 5.2 costs exactly twice as much, and its cache reads now cost the same $0.20 as Opus 5.5.
- No long-context surcharge. The pricing page says Claude 4.6 and later models (except Haiku 5.5) bill the full 1M context at standard rates. A 900K-token prompt costs the same per token as a 9K one.
- US-only inference costs 1.1x. Setting
inference_geo: "us"applies a 1.1x multiplier to every token category, fast mode included.
What does a real Fable 5.2 request cost?
We reuse the agentic coding turn from our Claude Ultra plan guide: 200K input tokens, 80% served from cache, plus 20K output tokens, with cache-write costs left out. This is our own arithmetic on list prices, not a measurement.
| Model | Fresh input (40K) | Cache reads (160K) | Output (20K) | Cost per turn | 1M in + 200K out, uncached |
|---|---|---|---|---|---|
| Fable 5.2 | $0.320 | $0.032 | $0.800 | $1.152 | $16.00 |
| Fable 5.1 | $0.400 | $0.040 | $1.000 | $1.440 | $20.00 |
| Opus 5.5 | $0.160 | $0.032 | $0.400 | $0.592 | $8.00 |
At the same token count, Fable 5.2 is exactly 20% cheaper than 5.1. The real saving per finished task could be larger or smaller. Thinking tokens are billed as output, and Anthropic’s prompting guide says that at high effort and above, Fable 5.2 thinks before almost every reply. Measure cost per completed task on your own workload, not cost per token. Our Fable pricing and limits breakdown walks through caching and batch break-even math in more detail.
What are Claude Fable 5.2’s context window and output limits?
| Spec | Claude Fable 5.2 |
|---|---|
| Claude API model ID | claude-fable-5-2 (pinned snapshot, no date suffix) |
| Context window | 1M tokens (default and maximum) |
| Max output | 128K tokens (thinking counts toward it) |
| Thinking | Adaptive, always on; depth set by effort |
| Effort levels | low, medium, high (API default), xhigh, max |
| Knowledge cutoff | June 2026 (reliable and training data) |
| Data retention | 30 days required; no zero data retention unless Anthropic authorizes it |
| Retirement | Not sooner than November 9, 2027 |
On the current tokenizer, Anthropic says 1M tokens is roughly 555,000 English words. Effort is the main cost lever. The docs recommend starting at high, stepping up to xhigh only for long coding and agent work where you have measured a gain, and keeping max for the rare task where time and tokens don’t matter.
How does Fable 5.2 score on benchmarks?
The table below comes from Anthropic’s launch post. These are vendor-reported results, run with safeguards at the highest available effort unless noted. Several of these benchmarks are new or internal, and we haven’t seen independent reproductions yet.
| Benchmark (area) | Fable 5.2 | Fable 5.1 | Opus 5.5 |
|---|---|---|---|
| FrontierCode 2.0 (agentic coding) | 49.1% (xhigh) | 32.4% (xhigh) | 45.5% |
| FrontierSWE v2 (agentic coding) | 69.9% | 56.3% | 62.3% |
| Terminal-Bench-Science 0.1 | 74.6% | 52.6% | 58.7% |
| OSWorld 2.1 (computer use) | 89.3% | 83.5% | 87.2% |
| Humanity’s Last Exam, with tools | 74.2% | 65.6% | 67.7% |
| Remote Labor Index (knowledge work) | 41.3% (high) | 17.9% | 21.3% |
| AA-Briefcase v1.1 (knowledge work) | 66.1% | 61.5% | 61.2% |
| AutomationBench (business workflows) | 51.4% | 31.4% | 42.5% |
| Chartography, no tools (visual) | 73.4% | 44.8% | 64.4% |
Fable 5.2 doesn’t win every row. On AutomationBench, Google’s Gemini 4 Carbon scores 53.9%, ahead of Fable 5.2. On Chartography, Fable 5.2 leads, but Gemini 4 Carbon (72.5%) and OpenAI’s GPT-6 Astra (71.0%) are close behind. Anthropic’s own wording is that Fable 5.2 is “a step above Opus 5.5 in almost all the capabilities we evaluated.”
The more useful claims for buyers are about cost per task, not top scores. Anthropic says that at high effort Fable 5.2 scores 45.3% on FrontierCode, level with Opus 5.5’s 45.5%, for roughly half the cost per task. It also says that on Humanity’s Last Exam, Fable 5.2 at low effort matches Opus 5.5’s best score for under a tenth of the cost. If those claims hold on your tasks, Fable 5.2 at lower effort can beat Opus 5.5 on price despite costing twice as much per token. That’s a hypothesis to test with your own evals, not a reason to switch blind.
Long-horizon demo. Anthropic says 1,000 Fable 5.2 agents finished a Lean formalization of Fermat’s Last Theorem in 30 hours with minimal intervention, while Opus 5.5 in the same setup completed 26% in 18 hours. The agents started from 60 statements, a sketch of the classical route and 47 lemmas already proved in an earlier formalization. It’s a striking demo, but it’s Anthropic’s own experiment.
Who can use Claude Fable 5.2, and on which plans?
Claude apps and Claude Code
The Help Center article “Claude Fable models on your plan” says Fable 5.2 is available on all paid plans and not on Free. How it is billed depends on the plan:
| Plan | Normal Fable 5.2 access | Launch offer (from Nov 9, 2026) |
|---|---|---|
| Free | Not available | None |
| Pro ($20/mo) and Team standard seats | Usage credits at API rates, not plan limits | Fable inside plan limits, up to 50% of weekly usage, through Nov 16, 2026 11:59 pm PT, plus one Fable limit reset (expires the same time) |
| Max ($100/mo) and Ultra ($200/mo) | Included, up to 50% of the weekly limit | 1 to 3 Fable limit resets by subscription length, valid until March 8, 2027 |
| Team premium seats, premium seat-based Enterprise seats | Included, up to 50% of the weekly limit | 1 to 3 resets (premium seats on legacy seat-based Enterprise) |
| Enterprise | Off until an admin enables it; usage-based Enterprise pays API rates | Not for self-serve, usage-based or standard legacy seats |
The launch offers article sets the reset count by tenure, measured from your first paid bill: one reset under six months, two for six to 12 months, three for over 12 months. A reset restores the Fable part of your weekly limit, and on plans with a 5-hour window it also refills that window. Billing must have started before November 9, 2026 at 10:00 am PT. You trigger a reset under Settings > Usage > Limit resets > “Reset for free.”
The 50% figure is a share of your existing weekly limit, not extra capacity, and Anthropic doesn’t publish how fast Fable drains it. Some news coverage says Pro includes Fable up to 50% of weekly limits. That’s the launch week only. From November 17, Pro goes back to usage credits. Our Ultra plan explainer covers how the weekly and 5-hour limits interact.
In Claude Code, Fable 5.2 needs version 2.1.321 or later. The fable alias now resolves to Fable 5.2 (except in Claude apps gateway sessions, where it still means Fable 5), and no Fable model is the default on any plan. Select it with /model fable or claude --model fable. If your seat bills Fable to credits, the picker shows “Requires usage credits.” For a full cost picture of the agent, see our Claude Code pricing guide.
API and cloud platforms
On the API, Fable 5.2 is open to all customers. Per the release notes it is available on:
- the Claude API and Claude Platform on AWS, as
claude-fable-5-2; - Amazon Bedrock on the
bedrock-runtimeendpoint only, asglobal.anthropic.claude-fable-5-2orus.anthropic.claude-fable-5-2; - Claude Models on Google Cloud, both Anthropic-managed and Google-managed;
- Claude in Microsoft Foundry, on Anthropic infrastructure.
Fast mode works on the Claude API, both Google Cloud options, and Foundry (as a separate claude-fable-5-2-fast deployment). On the API you set speed: "fast" with no beta header.
What is Claude Mythos 5.2, and who can access it?
Anthropic says Fable 5.2 and Mythos 5.2 are the same underlying model. The difference is the safeguards. Fable 5.2 blocks high-risk cyber work (penetration testing, exploit generation, binary vulnerability research) and most research-biology tasks. In the Claude apps, a blocked cyber query is answered by Opus 4.8 and a blocked biology query by Opus 5. Mythos 5.2 runs without those safeguards, at the same $8 / $40 price.
Access is limited to organizations verified through Anthropic’s programs: the Cyber Verification Program and the Life Sciences Verification Program. Claude Enterprise customers can use Claude Security, which runs on Mythos 5.2, without joining the Cyber Verification Program. Mythos 5.2 is on the Claude API (claude-mythos-5-2), Bedrock, both Google Cloud options and Foundry. Anthropic calls it its most capable model for cybersecurity engineering, and reports that on an internal set of 1,829 CVEs it found 10.2% more vulnerabilities than Opus 5.5. If you’re new to the Mythos line, our Mythos explainer covers its history.
Zero data retention. Both models require 30-day retention and are not available under ZDR unless Anthropic expressly authorizes it. Anthropic is piloting Enterprise Frontier Safeguards, which run safety monitoring on data stored in the customer’s own infrastructure so eligible customers can keep ZDR. It’s in private beta with Bridgewater, Salesforce and Stripe.
How do you migrate from Fable 5.1 to Fable 5.2?
The code change is one string: replace claude-fable-5-1 with claude-fable-5-2. If Claude Code, claude.ai, Claude Managed Agents or the Claude Agent SDK builds your requests, you’re done. If your own code builds the messages array, check four breaking changes first:
- The prefix check now applies to every account. If your code edits the system prompt, the tools or an earlier message and then sends back a Fable 5.2 thinking block, the request fails with a 400 (“The block is bound to a different conversation”). On 5.1 this only hit accounts created on or after August 31, 2026, so code that worked on an older account can break after the switch. Keep history append-only. The
drop_blocksetting under thethinking-binding-controls-2026-08-01beta is a stopgap, and it can raise output tokens. - Only Fable 5.2 and Mythos 5.2 read Fable 5.2 thinking blocks. If a router or fallback moves a conversation to Opus 5.5, Fable 5.1 or the refusal fallbacks (Opus 4.8, Opus 5), the reasoning from earlier turns is dropped. The request still succeeds, and dropped blocks aren’t billed.
computer_20251124returns a 400 on every platform except Amazon Bedrock. Use thecomputer_toolset_20260801toolset.- Advisor tool pairing. A Fable 5.2 executor accepts only Fable 5.2 or Mythos 5.2 as its advisor. Change the advisor model in new conversations, because changing
toolsmid-conversation trips the prefix check.
Rules carried over from 5.1: thinking: {"type": "disabled"} and budget_tokens return a 400, and so does forced tool_choice (any or tool). Use auto with strict: true, or structured outputs. Organizations with customer-managed encryption keys can’t use structured outputs on Fable 5.2.
Anthropic also lists behavior changes that don’t break code but can surprise you. Prompt lines like “answer directly without deliberating” no longer stop the model thinking at medium and above, so lower the effort instead. On long tasks at low or medium effort, a turn can end with a list of what remains rather than finished work. Re-run your evals with an effort sweep starting at high. If you’re also moving lower-tier traffic, our Sonnet 5.5 migration guide covers the equivalent changes on Sonnet.
Should you switch to Claude Fable 5.2?
- You run Fable 5.1 on the API: yes, after the four checks above. Same capabilities or better, 20% lower list price. Fable 5.1 stays active and won’t retire before September 1, 2027, so there’s no rush if your harness edits history.
- You run Opus 5.5: only when your evals show a gap. Anthropic still recommends starting with Opus 5.5 for most workloads and moving to Fable 5.2 when Opus 5.5 at higher effort falls short. Do test Fable 5.2 at
lowandmedium, given Anthropic’s cost-per-task claims. - You’re on Max or Ultra: pick it with
/model fable. It draws on the same Fable half of your weekly limit, and the launch resets add headroom until March 8, 2027. - You’re on Pro: try it free until November 16, then decide whether usage credits or the $80 step to Max makes more sense for your Fable volume.
- You need ZDR or computer use on
computer_20251124: stay put or plan the migration. Neither works on Fable 5.2 without changes.
Frequently asked questions
When was Claude Fable 5.2 released?
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026. Fable 5.2 succeeds Fable 5.1, which launched on September 1, 2026, and is available on the Claude API, the Claude apps and the major cloud platforms.
How much does Claude Fable 5.2 cost?
On the API, Fable 5.2 costs $8 per million input tokens and $40 per million output tokens, down from $10 and $50 on Fable 5.1. Cache reads cost $0.20, batch processing is $4 and $20, and optional fast mode is $16 and $80 per million tokens.
What is the Claude Fable 5.2 model ID?
The Claude API model ID is claude-fable-5-2, with no date suffix. On Amazon Bedrock it is global.anthropic.claude-fable-5-2 or us.anthropic.claude-fable-5-2 on the bedrock-runtime endpoint. Claude Mythos 5.2 uses claude-mythos-5-2.
What is the context window of Claude Fable 5.2?
Fable 5.2 has a 1M-token context window, which is both the default and the maximum, and up to 128K output tokens. The whole 1M context is billed at standard rates, with no long-context surcharge.
Is Claude Fable 5.2 included in Claude Pro?
Only during the launch week. Pro and Team standard seats get Fable inside plan limits, up to 50% of weekly usage, through November 16, 2026, plus one reset. After that, Fable on Pro runs on usage credits at API rates. Max, Ultra and Team premium seats include Fable up to 50% of the weekly limit.
What is the difference between Claude Fable 5.2 and Claude Mythos 5.2?
They are the same underlying model at the same price. Fable 5.2 has cyber and biology safeguards and is open to everyone. Mythos 5.2 runs without those safeguards and is only for organizations verified through Anthropic’s Cyber or Life Sciences Verification Programs.
Is Fable 5.2 better than Opus 5.5?
On Anthropic’s own benchmarks, Fable 5.2 beats Opus 5.5 on almost every test, but it costs twice as much per token. Anthropic still recommends starting with Opus 5.5 for most work and moving to Fable 5.2 when your evals on Opus 5.5 fall short.
Polish readers can find this guide in Claude Fable 5.2 po polsku.
Sources & methodology
- Anthropic — Introducing Claude Fable 5.2 and Claude Mythos 5.2 (November 9, 2026; edited November 11). Benchmarks, speed, safeguards, availability.
- Claude API Docs — Release notes, November 9, 2026 entry.
- Claude API Docs — What’s new in Claude Fable 5.2 (breaking changes, pricing, availability, migration).
- Claude API Docs — Models overview (specs, default effort, knowledge cutoff, retirement).
- Claude API Docs — Pricing (cache, batch, fast mode, data residency, long context).
- Claude API Docs — Model deprecations (Fable 5.2 and 5.1 lifecycle).
- Claude Help Center — Claude Fable models on your plan.
- Claude Help Center — Claude Fable 5.2 launch offers.
- Claude — Plans & pricing.
- Claude Code docs — Model configuration (fable alias, v2.1.321, default effort).
- 9to5Mac — Anthropic upgrades Claude with new Fable 5.2 model (November 9, 2026), secondary cross-check.
Sources accessed November 12, 2026. Prices, limits and availability come from Anthropic’s documentation and Help Center. Benchmark results and the cost-per-task comparisons are Anthropic’s own claims and have not been independently reproduced. Per-turn costs are DecodeTheFuture’s arithmetic on list prices for an illustrative token profile, not measurements.

