Overview
Anthropic upgraded Claude Fable and Claude Mythos on November 9, 2026, releasing Claude Fable 5.2 and Claude Mythos 5.2. Anthropic describes them as the same underlying model — Fable 5.2 has stronger cyber and biology safeguards, which allows it to be available to all users, while Mythos 5.2 is restricted to verified organizations. Both improve on coding, reasoning, and vision, generate outputs up to 30% faster than Claude Fable 5.1 and Claude Mythos 5.1, and are available at a lower price of $8 per million input tokens and $40 per million output tokens. The model IDs are claude-fable-5-2 and claude-mythos-5-2.
Key features
-
Faster outputs at a lower price
Claude Fable 5.1 cost $10 / $50 per MTok, and in long agentic sessions output speed adds up to wait time. Fable 5.2 and Mythos 5.2 generate outputs up to 30% faster than the 5.1 models, at $8 / $40 per MTok. On Medium effort (the default in Claude apps and Claude Code), Anthropic says Fable 5.2 can accomplish challenging tasks with fewer tokens than Fable 5.1.
-
Fast mode
Fable 5.2 and Mythos 5.2 also come with an optional fast mode, which speeds up token generation by up to 2.5 times, at $16 per million input tokens and $80 per million output tokens. Fast mode is available on Claude Code, the Claude Platform, Claude Models on Google Cloud, and Microsoft Foundry (for deployments hosted on Anthropic, as a separate model,
claude-fable-5-2-fast). Fast mode for Mythos 5.2 is limited to Verification Program organizations. -
Agentic coding
Over 90% of the changes merged into Anthropic’s core codebase are now written by Claude (with human guidance). Agents running an early Fable 5.2 build migrated a critical production service from Python to Rust, and the service now starts about 26 times faster. Starting from an empty folder, the model coordinated over 1,000 subagents working around the clock for eight days, and produced a working operating system that boots on a laptop.
-
Advisor and Orchestrator roles
Fable 5.2 can also play the role of Advisor (consulted by smaller models) or Orchestrator (where it delegates simpler tasks to smaller models). With the advisor tool (beta) on the API, every executor model accepts
claude-fable-5-2as its advisor. -
Science and mathematics
A team of 1,000 Fable 5.2 agents fully formalized Fermat’s Last Theorem in Lean in just 30 hours; an earlier formalization, performed by an unreleased research version of Claude, took 11 days. Claude Fable 5.2 solved two open problems in theoretical physics, both concerning black holes, and Mythos 5.2 developed a method for predicting full viral capsid structures, accurately modeling 28 of the 34 it attempted.
-
Cybersecurity (Mythos 5.2)
On Anthropic’s internal CVE Rediscovery evaluation, Claude Mythos 5.2 found 10.2% more vulnerabilities than Claude Opus 5.5 and 14.4% more than Claude Mythos 5.1. Claude Enterprise customers (even if they aren’t members of the Cyber Verification Program) can access Claude Security, which scans codebases for vulnerabilities and now includes Claude Mythos 5.2.
-
Fewer safeguard blocks
When a safeguard blocks a request, the work stops there. Anthropic observed about 55% fewer blocks with Fable 5.2 compared with Fable 5.1 — about 60% fewer in cybersecurity and 40% fewer in biology. Fable’s classifiers still block high-risk cyber activity such as penetration testing and exploit generation, and most research biology tasks; for those, users instead receive a response from Opus 4.8 for cybersecurity and Opus 5 for biology.
Benchmarks
| Benchmark | Fable 5.2 | Fable 5.1 | Opus 5.5 |
|---|---|---|---|
| FrontierCode 2.0 (Agentic coding) | 49.1% (Xhigh effort) | 32.4% (Xhigh effort) | 45.5% |
| FrontierSWE v2 (Agentic coding) | 69.9% | 56.3% | 62.3% |
| AutomationBench (Business workflows) | 51.4% | 31.4% | 42.5% |
| Terminal-Bench-Science 0.1 (Agentic scientific research) | 74.6% | 52.6% | 58.7% |
| WANDR, relaxed score (Agentic search) | 82.1% | 68.7% | 72.3% |
| OSWorld 2.1, offline set, partial (Computer use) | 89.3% | 83.5% | 87.2% |
| RoboDojo Sim (Robotics) | 64.9% | 27.6% | 39.4% |
| AA-Briefcase v1.1, rubric (Knowledge work) | 66.1% | 61.5% | 61.2% |
| AA-Briefcase v1.1, Elo (Knowledge work) | 1785 | 1669 | 1800 |
| Remote Labor Index (Knowledge work) | 41.3% (High effort) | 17.9% | 21.3% |
| Humanity’s Last Exam (with tools) | 74.2% | 65.6% | 67.7% |
| Chartography (Visual chart recognition, no tools) | 73.4% | 44.8% | 64.4% |
Mythos 5.2 scores 89.6% on CVE Rediscovery (Mythos 5.1: 78.3%; Opus 5.5 without safeguards: 81.3%) and 76.3% on CyScenarioBench (Mythos 5.1: 61.7%).
Pricing & availability
| Item | Detail |
|---|---|
| Input / output | $8 / $40 per MTok (Claude Fable 5.1: $10 / $50) |
| Cache reads | $0.20 per MTok (0.025x the base input price, same multiplier as 5.1) |
| Cache writes | $10 per MTok (5m), $16 per MTok (1h) |
| Batch processing | $4 input / $20 output per MTok |
| Fast mode | $16 input / $80 output per MTok |
| Model IDs | claude-fable-5-2, claude-mythos-5-2 |
| Context window / max output | 1M tokens / 128K tokens |
| API default effort | high |
| Platforms (Fable 5.2) | Claude API, Amazon Bedrock (anthropic.claude-fable-5-2), Claude Platform on AWS, Claude Models on Google Cloud (Anthropic managed and Google managed), Microsoft Foundry |
| Mythos 5.2 | Organizations verified through Anthropic’s verification programs, such as the Cyber Verification Program and the Life Sciences Verification Program |
Notes
- Four changes from Claude Fable 5.1 can break your code — First, the prefix check applies to every account: editing the
systemprompt, thetools, or an earlier message invalidates thinking blocks and returns a 400 error whatever the account’s age. On Claude Fable 5.1 this only applied to accounts created on or after August 31, 2026, so code that edits its history can run without errors on an older account before the switch and fail after it. Second, only Claude Fable 5.2 and Claude Mythos 5.2 read Claude Fable 5.2 thinking blocks, so moving to Claude Opus 5.5 or Claude Fable 5.1 loses reasoning for those turns. Third, the earliercomputer_20251124tool returns a 400invalid_request_erroron every platform except Amazon Bedrock; declarecomputer_toolset_20260801instead. Fourth, with the advisor tool, a Claude Fable 5.2 executor accepts onlyclaude-fable-5-2orclaude-mythos-5-2as its advisor. - Nothing changes if Claude Code or claude.ai builds your requests — the same holds for Claude Managed Agents and the Claude Agent SDK. If your own code builds the
messagesarray, run the test in the migration guide’s prefix check section before switching. - Default effort depends on where you use it — Medium is the default in Claude apps and Claude Code; the API default is
high. The documentation says to start athigh, step up toxhighfor your hardest long-running coding and agentic work, and reservemaxfor measuring the model’s ceiling. - Behavior differences — prompt lines that ask for less thinking don’t stop the model from thinking at
mediumeffort and above; a turn can end before the work is done on long, multistep tasks (more often atlowandmedium); and long runs of tool calls can pass with little user-facing text. “Prompting Claude Fable 5.2” has guidance for each. - Unchanged from Claude Fable 5.1 — adaptive thinking is always on and can’t be turned off, forced tool use (
tool_choicetypesanyandtool) returns a 400 error, andtemperature,top_p, andtop_kshould be omitted. - Refusals and billing — a declined request returns HTTP 200 with
stop_reason: "refusal". Whether a refusal that arrives before any output is billed depends on its refusal category, and it counts against your rate limits either way. - Plan usage — Fable is included for up to 50% of weekly usage limits for Pro plans and standard seats on the Team plan. As a launch promotion, plans and seats that include Fable models get one to three Fable resets based on how long they’ve subscribed.
- Data retention — both models require 30-day data retention and aren’t available under zero data retention unless expressly authorized by Anthropic. Per the announcement, eligible customers can continue using Claude Fable 5.2 and Claude Mythos 5.2 with zero data retention before implementing safety monitoring with EFS; other customers get 30-day retention for safety monitoring.