Claude Fable 5.2 API
NewLLMThe Claude Fable 5.2 API is Anthropic's newest Fable revision: 1M context, always-on adaptive thinking and a 128k output ceiling, with lower list rates than Fable 5.1 and faster answers.
- Input · USD / 1M tokens$2.8 USDOfficial price
$8 USD - Output · USD / 1M tokens$14 USDOfficial price
$40 USD - Savings65%
- Context1M
- View tool pricing and billing rules
What are we making today?
Bring an idea, a question, or a first draft.
01
What this model is for
The Claude Fable 5.2 API is Anthropic's newest Fable revision: 1M context, always-on adaptive thinking and a 128k output ceiling, with lower list rates than Fable 5.1 and faster answers.
Claude Fable 5.2 is Anthropic's Fable revision for work that runs for hours without a person watching, released on November 9, 2026. It keeps Fable 5.1's 1M-token window and always-on adaptive thinking, raises the maximum output to 128k tokens and sits at the top of the Fable tier in the catalog. Anthropic describes the launch as a cheaper and faster revision rather than a new architecture, and the model ships with vision and tool use enabled.
Anthropic still recommends starting most work on Opus 5.5 and moving to Fable when an evaluation says so. Treat the id swap as an experiment with a hypothesis: pick a task where Opus 5.5 stalls, record the failure, then replay the same prompt, tools and effort setting here. Adaptive thinking cannot be switched off and effort defaults to high, so the comparison is only fair when both runs use the same effort.
Best-fit coding workloads
Multi-hour repository work
Hand it a defect that crosses packages and let it read, patch and rerun tests across many passes without per-step approval. State the definition of done in the first message and ask for a short progress note at fixed intervals, so a run that drifts can be stopped early.
Long-horizon research
Give a question, the sources you trust and a citation format, then let it search, read and revise its outline across turns. Require a table that pairs each claim with its source, so a reviewer checks the trail instead of rereading every page the run touched.
Large-context migrations
Feed a whole service, its tests and the target API notes into one request and ask for a patch plus a rollback plan. The 128k output ceiling is what lets a multi-file answer fit in a single turn; review the change list before applying it.
02
Hands-on notes
Checked on 2026-11-09 through the OmniaKey API with one standard coding prompt (median of 3 runs) and three capability probes.
- Output tokens for our standard coding prompt, reasoning included
- 1,362
- Cost of that output at current prices
- $0.0191
- JSON mode
- Passed
- Tool calling
- Passed
- Image input
- Passed
Anthropic's launch notes carry four migration checks: keep the conversation prefix stable, leave adaptive thinking on, move computer-use agents to the newer toolset, and confirm the 30-day retention requirement. Our launch breakdown walks each one before you swap the id.
Read the testIn our own smoke run the id answered every request on the first attempt, JSON mode and tool calls came back in the documented shape, and an image sent inline as a data URL was read correctly, so switching ids did not change what the gateway returns.
03
Technical specification
Catalog facts and integration details verified for this route.
- Model ID
- claude-fable-5-2
- Context
- 1M
- Inputs and tools
- Vision · Tool use · Adaptive thinking
- API endpoint
- /v1/chat/completions
- Cache hit / 1M
- $0.07 USD
- Coding clients
- Claude Code · Cursor · Cline · Aider
Limits to plan around
- Editing the system prompt, the tool list or an earlier message and then reusing a later thinking block returns an HTTP 400 unless the migration option is set. When you need a different prefix, start a fresh conversation instead of trimming history.
- Thinking cannot be turned off and effort defaults to high, so short prompts pay for reasoning they may not need. Test a lower effort on your own task before assuming the default is the cheapest setting.
- The route requires 30-day data retention, and zero-retention use needs Anthropic's explicit permission. Confirm the policy before sending source code, incident logs or personal data.
- Computer-use agents must switch to the newer toolset definition; the previous one is not accepted on this model outside Bedrock.
04
Compare nearby models
Compare workload fit and current listed token rates before choosing a route.
| Model | Best for | Context | Input | Output |
|---|---|---|---|---|
| Claude Fable 5.2This page claude-fable-5-2 | Long autonomous runs that want the newest Fable revision and a bigger output ceiling. | 1M | $2.8 USD Official price $8 USD | $14 USD Official price $40 USD |
| Claude Fable 5.1 claude-fable-5-1 | Runs that already work, where a prompt replay has not yet shown a difference. | 1M | $2.8 USD Official price $10 USD | $14 USD Official price $50 USD |
| Claude Opus 5.5 claude-opus-5-5 | Almost every first attempt: cheaper to evaluate and the starting point Anthropic recommends. | 1M | $1.12 USD Official price $4 USD | $5.6 USD Official price $20 USD |
| Claude Opus 5 claude-opus-5 | Agents pinned to Opus 5 whose prompts and tool schemas have not been replayed on 5.5. | 1M | $1.4 USD Official price $5 USD | $7 USD Official price $25 USD |
05
SDK quickstarts
Use one base URL and model ID from cURL, Python, or TypeScript.
curl https://api.omniakey.com/v1/chat/completions \
-H "Authorization: Bearer $OMNIAKEY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-fable-5-2",
"messages": [
{"role": "user", "content": "Review this pull request"}
],
"stream": true
}'06
Estimate token cost
Enter expected input and output tokens to compare current listed rates.
- OmniaKey estimate
- $0.56
- Official-rate estimate
- $1.60
- Estimated savings
- $1.04
Estimate covers input and output tokens only. Cache usage, retries, tools, and client-side token accounting can change the final charge.
07
Related coding guides
Deeper selection, setup, and troubleshooting guidance for this model family.
Claude Fable 5.2: New Features, Pricing and API
Fable 5.2 cuts standard token prices 20% and adds a faster mode, with a 1M context window and 128k output.
Read guideClaude Fable 5.1 Review
Claude Fable 5.1 is Anthropic's newest long-horizon model. Here are its features, Fable 5 differences, API price changes, and compatibility notes.
Read guideClaude Fable 5 Pricing and API Access
Anthropic lists Fable 5 at $10 input, $1 cache read, and $50 output per million tokens.
Read guideClaude Fable 5.5 release status
What is confirmed and which Claude model to use while you wait.
Read guide08
Sources and verification
Model facts and links were checked on 2026-11-09. Provider documentation remains the authority for upstream behavior.
Frequently asked questions
How much does Claude Fable 5.2 cost on OmniaKey?
OmniaKey currently charges $2.8 USD per 1M input tokens and $14 USD per 1M output tokens, in USD. Direct provider rates are separate and may change with input context length.
What is the context window for Claude Fable 5.2?
The listed context window is 1M. Actual usable limits can also depend on the client and upstream request format.
What is Claude Fable 5.2?
It is Anthropic's November 2026 Fable revision: a 1M-token context window, 128k maximum output, always-on adaptive thinking, vision and tool use, and lower list rates than Fable 5.1. Anthropic positions it for demanding reasoning and long-horizon agentic work rather than everyday tasks.
How is Fable 5.2 different from Fable 5.1?
The window and the thinking behaviour are the same; the output ceiling doubles to 128k, Anthropic reports faster generation and lower standard rates, and the computer-use toolset changes. Existing Claude Code and Agent SDK clients do not need request changes.
Do I need to change my client to use Fable 5.2?
Only the model id. Custom clients should keep the conversation prefix stable, leave adaptive thinking on, update computer-use tool definitions and confirm the retention policy. Editing earlier turns and then reusing later thinking blocks is what fails, not the id swap itself.
Does Fable 5.2 still accept images and tool calls?
Yes. In our smoke run the model read an image sent as a data URL and returned a well-formed tool call, and the same id answers on both the OpenAI-compatible POST /v1/chat/completions and the native POST /v1/messages route through OmniaKey.