Claude Fable 5.2 API Pricing

Anthropic's pricing table lists $8/M base input, $10.00/M 5-minute cache writes, $16/M 1-hour cache writes, $0.20/M cache hits and $40/M output. Batch is half of base at $4/M and $20/M. Under this site's 92/8 agentic blend at an 82% cache-hit rate the effective rate is $5.01/M.

Every price moved by the same factor. Base input went from $10 to $8, output from $50 to $40, the two cache-write tiers from $12.50 and $20 to $10 and $16, the cache hit from $0.25 to $0.20, and batch from $5/$25 to $4/$20. Because the whole card is 0.8x Fable 5.1, the saving is 20% for any token mix, any cache-hit rate and batch or not: our blend goes from $6.26/M to $5.01/M. The cache hit stays at 0.025x base input, the multiplier Anthropic introduced with the 5.1 pair. Anthropic also says Fable 5.2 generates output up to 30% faster than 5.1.

$16/M input and $80/M output, twice the standard rates, applied across the full context window. Anthropic says it speeds up token generation by up to 2.5 times. It runs on the Claude API, on Claude Models on Google Cloud (both the Anthropic-managed and the Google-managed offering, where Google sets its own price) and on Microsoft Foundry for Anthropic-hosted deployments. Fast mode does not work with the Batch API.

Opus 5.5 has half the base card at $4/$20, but the same $0.20/M cache hit, so on our blend it costs $2.58/M against Fable 5.2's $5.01/M. Anthropic's advice is to start with Opus 5.5 and move up to Fable for the most ambitious, open-ended technical work. Per-token price is not the whole story, though: Anthropic reports that on FrontierCode 2.0, Fable 5.2 at High effort scored about level with Opus 5.5 at Max effort (45.3% against 45.5%) for roughly half the cost per task, because it needed fewer tokens. That is the vendor's own test; check it on your workload.

Anthropic says they are the same underlying model with identical rate cards. Fable 5.2 carries strong cybersecurity and biology safeguards, which is why anyone can call it. Mythos 5.2 is for organisations verified through Anthropic's Cyber and Life Sciences Verification Programs, and there is no self-serve sign-up.

No. Anthropic's pricing page says Claude 4.6 and later models include the full 1M-token window at standard pricing, so a 900k-token request is billed at the same per-token rate as a 9k-token one, and caching and batch discounts apply across the whole window. The modifier that does cost more is pinning inference to the US with inference_geo: a 1.1x multiplier on every token category.

Not by default. The model page says Claude Fable 5.2 requires 30-day data retention and is not available under zero data retention unless Anthropic expressly authorizes it. If your contract depends on ZDR, check that before you migrate from an Opus or Sonnet model.

Not directly. Anthropic notes that Claude 4.7 and later models use a newer tokenizer producing roughly 30% more tokens for the same text, with the exact increase depending on content. Fable 5.2 and Fable 5.1 share that tokenizer, so the 20% cut between them is like for like. Against a model on the older tokenizer, the same nominal $/M buys less real text.