Based on each vendor’s official docs, checked November 11, 2026

Claude Fable 5.2: model choice, CursorBench and Pro week

Claude Fable 5.2 is Anthropic’s most capable generally available Claude model, released November 9, 2026, for demanding reasoning and long-horizon agentic work. Anthropic’s launch post and model docs, the official choosing-a-model guide, the Fable 5.2 launch-offers article, Cursor’s Fable 5.2 model page and the CursorBench 4.0 leaderboard were all checked November 11, 2026. Figures below are those vendors’ published numbers, not independent measurements and not a promise about your project. Fable and Mythos 5.2 are the same underlying model; Mythos is limited to organizations in Anthropic’s verification programs.

1. 1. Start from the published specs and price

   On the Claude API, the model ID is claude-fable-5-2. Anthropic’s Fable 5.2 overview lists a 1M-token context window, 128K max output, always-on adaptive thinking, and a default effort of high. Base API pricing is $8 per million input tokens and $40 per million output tokens. An optional fast mode is $16 per million input tokens and $80 per million output tokens, and Anthropic says it can speed up token generation by up to 2.5 times. Prompt-cache reads are $0.20 per million tokens; a 5-minute cache write is $10 and a 1-hour cache write is $16. The Batch API is 50% off input and output. Anthropic also says Fable 5.2 and Mythos 5.2 generate outputs up to 30% faster than the 5.1 versions. These are list prices and vendor claims from the pages checked today, not a bill for a specific client or Cursor plan.

   claude --model fable
   # or, in a session:
   /model fable

2. 2. Choose Fable only when Opus 5.5 is not enough

   Anthropic’s choosing-a-model guide says most workloads should start with Claude Opus 5.5, described as the strongest daily driver for complex agentic coding and enterprise work. Step up to Fable 5.2 when you need the highest available capability, such as agent sessions that run for hours, multistep deep research, or analysis carried through to a finished document, spreadsheet or deck, or when your own evals at xhigh or max effort on Opus 5.5 still fall short on demanding reasoning or long-horizon agentic work. For everyday coding, Anthropic points to Claude Sonnet 5.5; for the lowest latency and price, including subagent tasks, it points to Claude Haiku 5.5. Fable 5.2 can also act as an Advisor that smaller models consult, or as an Orchestrator that delegates simpler work. Do not default to Fable for every task just because it is newer.

3. 3. Tune effort before you switch models

   Anthropic says tuning effort is often a better lever than switching models. On Fable 5.2 the API default is high; start there and adjust up or down against your own evals. On Opus 5.5 and Haiku 5.5 the default is medium. Cursor’s Fable 5.2 docs use the same idea: start with high, and do not carry over a Fable 5.1 effort setting because the levels behave differently. Cursor lists effort as low, medium, high, extra high and max, and says to save extra high and max for the hardest long-running tasks. Anthropic’s Fable 5.2 prompting guide similarly reserves xhigh and max for work where you have measured a gain; max is for measuring the ceiling or rare tasks where time and tokens do not matter, and those turns are slow and use many more tokens.

4. 4. Select Fable explicitly in Claude Code

   Claude Code’s model-config docs say no Fable model is the account-type default on any plan or provider; you have to select it. Use /model fable or launch with claude --model fable. Unless you set ANTHROPIC_DEFAULT_FABLE_MODEL, the fable alias resolves to Fable 5.2 except in Claude apps gateway sessions, where it resolves to Fable 5 and you should select claude-fable-5-2 by ID if the gateway serves it. Fable 5.2 requires Claude Code v2.1.321 or later. Depending on your plan and seat, Fable usage can bill to usage credits instead of included limits; the /model picker may show “Requires usage credits.” Anthropic’s launch-offers article also notes that your default model does not change during the Pro week: pick a Fable model from the model picker to use it. Requests that Fable’s safety classifiers flag, most often in cybersecurity and biology, trigger automatic model fallback.

5. 5. Use the launch offers as written, not as unlimited free usage

   For Pro plans and standard seats on Team plans, Anthropic includes all Fable models in the plan’s weekly usage limits from launch on November 9, 2026 through November 16, 2026 at 11:59 pm PT, up to 50% of weekly usage limits, plus one Fable limit reset if paid billing was in place before November 9, 2026 at 10:00 am PT. After you use that 50%, or once the week ends, Fable usage is billed to usage credits if they are turned on. The Pro and standard Team reset expires at the same November 16 deadline. Max plans, Ultra plans, premium Team seats and premium seats on legacy seat-based Enterprise plans do not get that included week; they get one to three Fable limit resets based on subscription length (under six months: one; six to 12 months: two; more than 12 months: three), expiring March 8, 2027 at 11:59 pm PT. The offers are not available on the Free plan, standard seats on legacy seat-based Enterprise, self-serve or usage-based Enterprise, or HIPAA-ready Enterprise. A reset restores Fable usage on the weekly limit and, if the account has a five-hour limit, refills that window; it does not raise your limits. Use a reset from Settings > Usage. Anthropic warns that Fable models use usage limits faster than other Claude models.

CursorBench 4.0 scores, from Cursor’s leaderboard

Cursor’s CursorBench 4.0 page, checked November 11, 2026, evaluates agents on ambiguous, multi-file tasks from real Cursor sessions. Higher scores are better. On that leaderboard, Gemini 4 Carbon High is first at 60.4% ($9.22 cost per task) and Carbon Medium is second at 60.1% ($8.49). Fable 5.2 Max is third at 60.0% ($32.79). Fable 5.2 High is sixth at 56.9% ($9.70). Compare like with like: Fable 5.1 Max is 51.8% and Fable 5.1 High is 49.2%. Cursor notes that small score differences may not be statistically meaningful, and that average cost per task applies each model’s published per-million-token pricing to the tokens used. These are Cursor’s own benchmark results, not an independent audit, and they are not a forecast of your repo’s outcome.

Cursor’s Fable 5.2 model docs say the model replaces Fable 5.1 in Cursor, costs 20% less on input and output than Fable 5.1, and bills at the same $8 / $40 API rates from prepaid credit on The X Subscription (or the Other Models pool on legacy plans). It is a third-party model, so it is not part of the usage included with Cursor plans. If Privacy Mode is on, or you are on Enterprise, an admin must approve Fable 5.2’s Anthropic data-retention policy before the model is available. Cursor also routes a request that trips Fable’s guardrails to Claude Opus automatically. Input and output rates are described as about 2x Claude Opus 5.5.

What the launch examples do and do not mean

Anthropic’s announcement describes prerelease and internal work: a team of about 1,000 Fable 5.2 agents formalized Fermat’s Last Theorem in 30 hours in Lean, with Anthropic reporting that Opus 5.5 completed 26% in 18 hours in the same 1,000-agent setup; Fable 5.2 designed three electronic devices (a voice-assistant hub, a USB speakerphone and a handheld virtual pet) that Anthropic says worked without electrical modifications after a partner built the boards; and it reports other research, robotics and coding examples. Those are vendor-reported results from the launch post, not a guarantee that the same method will work for your task, and they are not a substitute for your own evals.

Fable 5.2 requires 30-day data retention on Anthropic’s side and, per the API overview, is not available under zero data retention unless Anthropic expressly authorizes it. Retirement is listed as not sooner than November 9, 2027. For current plan prices and what is included after the launch window, use the Claude Code pricing page on this site and Anthropic’s official plan articles.

Tips

• Start on Opus 5.5 for most coding work; move to Fable 5.2 when your evals at high effort still fall short.
• In Claude Code, select /model fable; the default model does not switch by itself, including during the Pro week.
• Upgrade Claude Code to v2.1.321 or later before expecting Fable 5.2 to run.
• On Cursor, check prepaid credit and Privacy Mode approval; Fable 5.2 is not included in Cursor plan usage.
• Treat CursorBench and Anthropic launch examples as vendor-reported, and compare scores at the same effort level.

Next steps

Sources

• https://www.anthropic.com/claude-fable-and-mythos-5-2 · checked 2026-11-11
• https://platform.claude.com/docs/en/models/fable-5-2/overview · checked 2026-11-11
• https://platform.claude.com/docs/en/about-claude/models/choosing-a-model · checked 2026-11-11
• https://platform.claude.com/docs/en/models/overview · checked 2026-11-11
• https://support.claude.com/en/articles/17506494-claude-fable-5-2-launch-offers · checked 2026-11-11
• https://code.claude.com/docs/en/model-config · checked 2026-11-11
• https://cursor.com/docs/models/claude-fable-5-2 · checked 2026-11-11
• https://cursor.com/cursorbench · checked 2026-11-11
• https://platform.claude.com/docs/en/build-with-claude/fast-mode · checked 2026-11-11
• https://platform.claude.com/docs/en/about-claude/pricing · checked 2026-11-11