Claude Fable 5.2: what's new, benchmarks, pricing and how it compares with Opus 5.5

Claude Fable 5.2 at a glance

What's new in Fable 5.2

• Better at long, hard technical work: Anthropic says it can run ambiguous engineering tasks for days and coordinate hundreds, even more than a thousand, subagents.
• Cheaper: input and output cost 20% less than Fable 5.1, and batch processing is half price ($4 and $20).
• Faster: outputs up to 30% faster than Fable 5.1, and a fast mode that's up to 2.5x faster again, at double the price.
• Robotics: Anthropic reports much better control of physical robots, including on tasks it wasn't trained for.
• Safeguards: cybersecurity and biology safeguards are built in. Flagged cyber queries are routed to Claude Opus 4.8 and flagged biology queries to Claude Opus 5.
• Data retention: Fable models require 30-day retention for safety monitoring by default; zero data retention needs Anthropic's authorization.

Claude Fable 5.2 benchmarks

From Anthropic's Fable page, with safeguards on unless noted. These are Anthropic's own numbers, not independent tests.

Claude Fable 5.2 pricing

API prices per million tokens, from Anthropic's docs on November 10, 2026. On Claude plans, Fable 5.2 comes with your subscription's usage: on Pro and standard Team seats it can use up to 50% of your weekly limits. Fast mode in Claude Code is billed from usage credits.

Fable 5.2 vs Opus 5.5: which should you use?

Anthropic's own advice is to start with Opus 5.5 for most work and move to Fable 5.2 when Opus 5.5 at higher effort still falls short.

• Price: Fable 5.2 costs twice as much per token ($8 and $40 against $4 and $20), and its default effort is high where Opus 5.5's is medium, so it also tends to think more per request.
• Capability: on Anthropic's tables Fable 5.2 leads Opus 5.5 on agentic coding (69.9% vs 62.3% on FrontierSWE v2), science and long-horizon work; the gap is smaller on computer use (89.3% vs 87.2% on OSWorld 2.1).
• Cost per task: Anthropic says Fable 5.2 at high effort scored about level with Opus 5.5 at max on FrontierCode 2.0 (45.3% vs 45.5%) for roughly half the cost per task, so for hard tasks it isn't necessarily the more expensive choice.
• Switching mid-conversation: only Fable 5.2 and Mythos 5.2 can read Fable 5.2's thinking, so moving a conversation from Fable 5.2 to Opus 5.5 loses its reasoning for the turns that run there.
• Rule of thumb: Opus 5.5 as the daily driver; Fable 5.2 for long-running agent jobs, hard bugs and work that fails on Opus.

Fable 5.2 vs Mythos 5.2, and vs Fable 5.1

• Mythos 5.2 is the same model with fewer restrictions on cybersecurity and biology work. It's only for organizations verified through Anthropic's programs, such as the Cyber Verification Program, and costs the same.
• On Anthropic's cyber and biology tables Mythos 5.2 scores higher than Fable 5.2 with its safeguards on; for everything else they're the same model.
• Against Fable 5.1, Fable 5.2 is cheaper, up to 30% faster, and scores higher across Anthropic's tables, for example 49.1% vs 32.4% on FrontierCode 2.0 and 41.3% vs 17.9% on the Remote Labor Index.

How to use Claude Fable 5.2 in Claude Code

1. Update and pick the model

   Update Claude Code, then run /model and choose Claude Fable 5.2, or start a session with --model claude-fable-5-2. On a Claude plan, it needs Pro, Max, Ultra, Team or Enterprise.

   # Update Claude Code, then pick the model for this session
   claude update
   /model                       # choose Claude Fable 5.2
   
   # Or start a session on it directly
   claude --model claude-fable-5-2
   
   # Fast mode on Fable 5.2 (Claude Code v2.1.321 or later)
   /fast

2. Leave effort at high to start

   Anthropic says most of the gain is already there at high. Step up to xhigh for your hardest long-running coding work once you've seen it help; keep max for rare cases where time and tokens don't matter.

3. Turn on fast mode when speed matters

   Run /fast while on Fable 5.2 (Claude Code v2.1.321 or later). It's up to 2.5x faster at double the price; on subscription plans it's paid from usage credits, not your plan's limits, and it isn't available on Bedrock, Vertex or Foundry.

4. Calling the API? Check the breaking changes

   Adaptive thinking is always on (budget_tokens and disabled return 400), forced tool use isn't supported, editing earlier turns invalidates thinking blocks, and the old computer_20251124 tool is replaced by the computer use toolset. Claude Code and the Agent SDK handle this for you.

   # Claude API: change the model ID
   model = "claude-fable-5-1"  # before
   model = "claude-fable-5-2"  # after
   
   # Adaptive thinking is always on; steer depth with effort
   # low | medium | high (default) | xhigh | max

Source: Anthropic's docs, What's new in Claude Fable 5.2

Give Fable 5.2 a long-running job

Fable 5.2 is built for work that runs for hours. Crewmark turns a Claude Code session into an AI employee with a role, a goal, working hours and a heartbeat; the employee uses whatever model its session runs on, so you can put Fable 5.2 on the hardest jobs and keep Opus 5.5 for the rest.

1. Pick a role

   Such as the AI software engineer for a backlog of hard issues.

2. Set the model in its session

   Run /model in the employee's Claude Code conversation and choose Fable 5.2.

3. Let it run

   It wakes on its heartbeat and works through the board; see how to run Claude Code 24/7.

FAQ