No Zero Data Retention for Claude Fable 5.2 on AI Gateway

Vercel on November 9 added Anthropic’s Claude Fable 5.2 to AI Gateway. Anthropic doesn’t offer Zero Data Retention for the model, so if your team has ZDR switched on, the gateway returns an error instead of sending the request, says Vercel’s changelog. Teams that want to use Fable 5.2 have to accept that prompts and completions are kept for 30 days. They are not used to train Claude.
Fable 5.2 also ships with cybersecurity and biology safety classifiers turned on, so some requests will be declined. To keep those requests served, add a models array to providerOptions.gateway. When the cybersecurity classifier declines a request, AI Gateway tries each listed model in order. Vercel’s example falls back to Opus 5.5, then Sonnet 5.5. The option works the same way in every API format the gateway supports.
That fallback lands on the model Anthropic itself tells people to start with. On its Fable and Mythos 5.2 page, Anthropic calls Opus 5.5 its “strongest daily driver” for most coding and knowledge tasks and suggests moving up to Fable or Mythos from there. Its release notes say Opus 5.5, launched September 22, “performs at the level of Claude Fable 5.1 on most work.”
Anthropic says Fable 5.2’s cyber and biology safeguards are as robust as those in Fable 5.1 but cause fewer unnecessary blocks. In September, Anthropic said Fable 5.1 had already cut cyber interventions by about 60% per session compared with Fable 5. Fable and Mythos are the same underlying model. The safeguards are what let Anthropic open Fable to all users. Organizations doing beneficial work in cybersecurity or biomedical research can apply for Mythos through Anthropic’s Life Sciences and Cyber Verification Programs.
AI Gateway charges the provider’s price with no markup and no platform fee on inference, including Bring Your Own Key (BYOK) requests. That means Fable 5.2 costs the same on the gateway as Anthropic’s own rate: $8 per million input tokens and $40 per million output tokens, which Anthropic says is lower than Fable 5.1’s price.
Vercel says the model is built for open-ended technical work and does better than earlier Fable models at coding, reasoning and vision. It can run long, ambiguous tasks unattended. It can also work as an orchestrator that hands simpler tasks to smaller models, or as an advisor that other Claude models consult on harder problems. Anthropic says it generates output up to 30% faster than Fable 5.1. Effort goes from low to max, and the API default is high.
The model name is anthropic/claude-fable-5.2. You can call it through the AI SDK, the Chat Completions, Responses and Anthropic Messages APIs, or a coding agent connected to the gateway. To use it in Claude Code, Codex or Cursor, install the latest Vercel CLI and run npx vercel ai-gateway setup.
Up next
Claude Code makes Fable 5.2 default at $8/$40 per Mtok
“Secrets such as refresh tokens” could reach another host when an MCP server redirected Claude Code’s OAuth requests, and version 2.1.321 is the release that fixes it. Have you ever checked where your MCP servers send your OAuth tokens?
Related stories
- Claude Fable 5.2 drops to $8/$40 and runs up to 30% faster
- Claude Fable 5.2 draws its sprites in raw JavaScript
- Cache reads get 75% cheaper on Claude Fable 5.1
- Fable 5.2 matched Opus 5.5 on nine tasks at almost 4x cost
- Linear and Ramp users can pay for Claude without an API key
- Gemini Decisions API scores 1,000 questions in one call
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

