Claude Fable 5.2 costs $8 in and $40 out, 20% under Fable 5.1 on every line. A coding-agent step drops from $0.40 to $0.32, still twice Opus 5.5.
Anthropic launched Claude Fable 5.2 today at $8 per million input tokens and $40 per million output. Fable 5.1 stays on the price list at $10 and $50, so the new flagship is the cheaper of the two on every line, cache reads and batch included.
Photo by MountainAsh on Unsplash
Short answer
Claude Fable 5.2 costs $8 per 1M input tokens and $40 per 1M output tokens. Cache hits are $0.20, cache writes $10 (5 minutes) and $16 (1 hour), Batch is $4/$20 and fast mode is $16/$80. That is 20% under Fable 5.1 on every line. A 30K-in, 2K-out coding-agent step costs $0.32, against $0.40 on Fable 5.1 and $0.16 on Opus 5.5. Mythos 5.2 has the same card for verified organisations only.
How much does Claude Fable 5.2 cost?
Every figure below is from Anthropic's pricing page, read on November 9, 2026, per million tokens.
| Line | Fable 5.2 | Fable 5.1 | Change |
|---|---|---|---|
| Input | $8 | $10 | -20% |
| Output | $40 | $50 | -20% |
| Cache write, 5 min | $10 | $12.50 | -20% |
| Cache write, 1 hour | $16 | $20 | -20% |
| Cache hit | $0.20 | $0.25 | -20% |
| Batch input / output | $4 / $20 | $5 / $25 | -20% |
Both models keep the 0.025x cache-read multiplier that only the Fable and Mythos 5.x rows use. The rest of the lineup reads cache at 0.05x (Opus 5.5, Sonnet 5.5) or 0.1x. One result of that: a cached token on Fable 5.2 now costs $0.20, the same as a cached token on Opus 5.5, a model with half its input price.
Fable 5.2 on the Batch API costs $4/$20, which is Opus 5.5's standard card to the cent. Context is 1M tokens with 128K max output, and adaptive thinking is always on, per the release notes.
What a request costs on Fable 5.2 vs the rest of the lineup
These are the three fixed workloads every model page on this site uses, priced at standard rates with no caching. All four models use Anthropic's newer tokenizer (Claude 4.7 and later), so the same text is the same token count on each and the comparison is like for like.
| Model | Chat reply | RAG answer over retrieved docs | Coding-agent step |
|---|---|---|---|
| Claude Fable 5.2 | $0.028 | $0.088 | $0.320 |
| Claude Fable 5.1 | $0.035 | $0.110 | $0.400 |
| Claude Opus 5.5 | $0.014 | $0.044 | $0.160 |
| Claude Sonnet 5.5 | $0.0070 | $0.022 | $0.080 |
Agent loops re-send most of their context, so caching changes the gap. Take the same coding-agent step with 27K of the 30K input tokens served from cache: Fable 5.2 costs about $0.109, Fable 5.1 $0.137, Opus 5.5 $0.057 and Sonnet 5.5 $0.029. Output is most of the bill there, which is why Fable stays about twice the price of Opus 5.5 even with cache reads at the same rate.
At 10,000 uncached agent steps a month that is $3,200 on Fable 5.2 against $4,000 on Fable 5.1. If you are on 5.1 and your evals pass on 5.2, the move is a straight 20% cut. If you moved to Opus 5.5 for price, nothing here changes that: it is still half the cost per step.
Fast mode: Fable 5.2 at $16/$80, and Opus 4.8 fast ends November 30
Fable 5.2 has fast mode on the Claude API at $16 input and $80 output, exactly twice its standard rate. Fast mode does not combine with Batch. The same pricing page lists Opus 5.5 fast at $8/$40 and Opus 5 and Opus 4.8 fast at $10/$50.
Anthropic deprecated fast mode for Claude Opus 4.8 on October 30, with removal on November 30, 2026. After that, a request to claude-opus-4-8 with speed: "fast" returns an error, and the release notes point to Opus 5.5. Per agent step that move is cheaper:
| Fast mode | Input / output per 1M | Coding-agent step (30K in, 2K out) |
|---|---|---|
| Opus 4.8 fast (removed Nov 30) | $10 / $50 | $0.40 |
| Opus 5.5 fast | $8 / $40 | $0.32 |
| Fable 5.2 fast | $16 / $80 | $0.64 |
What changed on November 9
- Claude Fable 5.2 (
claude-fable-5-2) and Claude Mythos 5.2 (claude-mythos-5-2) launched at $8/$40. Fable 5.1 and Mythos 5.1 stay listed at $10/$50. - Fable 5.2 is on the Claude API, Claude Platform on AWS, Amazon Bedrock, both Google Cloud options and Microsoft Foundry. Mythos 5.2 is limited to organisations verified through programs such as the Cyber Verification Program.
- Fable 5.2 enforces the prompt prefix check on every account: editing the system prompt, tools or an earlier message and then sending back a later Fable 5.2 thinking block returns a 400.
- Computer use needs the
computer_toolset_20260801toolset. The oldercomputer_20251124tool returns a 400 everywhere except Amazon Bedrock.
Sources, all checked November 9, 2026: Claude Platform release notes (November 9 and October 30 entries) and Anthropic pricing (model table, batch table and fast mode table).