claude-fable-5-1
For demanding reasoning and long-horizon agentic work
Modalities
In / Out Price
10% off$9/ 1M
$10 Official price
$45/ 1M
$50 Official price
Context
1M
Released
Sep 1, 2026
API call examples
POST /v1/chat/completionscurl --request POST 'https://api.tokatlas.com/v1/chat/completions' \
--header "Authorization: Bearer $UNIROUTER_API_KEY" \
--header 'Content-Type: application/json' \
--data '{
"model": "claude-fable-5-1",
"messages": [
{
"role": "user",
"content": "Explain quantum entanglement."
}
],
"temperature": 0.7
}'Providers & routes
Current provider · Anthropic
Platform configured price · Group: ccmax
USD / 1M tokens. Each row shows input and output from the same tier. Displayed rates use the lowest available group multiplier; your API key group may differ.
Claude Fable 5.1 for long-running work and version comparisons
Use it to turn open-ended investigations into reviewable documents, especially when several sources agree on some points and conflict on others.
Phased tasks
Track progress through specification, implementation, and acceptance.
Large repositories
Provide relevant modules and failing tests so the model can trace dependencies.
Long context and output
Keep technical specs and execution history in the same working set.
Adaptive reasoning
Compare a newer version with the same tools and checks to see whether it reduces rework.
About
Inputs for long-running tasks
Fable 5.1 provides a million-token window for repositories, specifications, and tool results, with up to 128K output. Adaptive reasoning stays on and defaults to high. Structure implementation, tests, and handoff notes for phase-by-phase acceptance.
Comparison with earlier Fable and Opus
Use existing Fable 5 or Opus results as a baseline. Evaluate research and broad code changes separately for traceable evidence, test results, and scope control across turns.
Pre-rollout checks for outputs and tools
Include known failure cases and check stopping and recovery. Verify the effort settings and tool structure on the selected endpoint before migrating long-running jobs.
A long software task can test planning, patching, and regression together. For migration documents or research, focus on evidence, decisions left for reviewers, and whether handoff can continue the work. Keep earlier-model runs to identify which phase improved.
claude-fable-5-1 FAQ
What should I retest when moving from Fable 5?
Retest long jobs, tool selection and reasoning across turns. Also compare actual cache costs when your workflow sends the same reference material repeatedly.
How can I compare Fable 5.1 with the model I already use?
Run the same task with the same tools and acceptance criteria, then record manual fixes and time to completion.
Do I need to resend the entire history after splitting a long task?
Save a summary, completed steps and open issues at each checkpoint. Carry forward the relevant files and tool results, then track the full task's acceptance result.
Can an API key be used to download and run Fable 5.1 locally?
An API key gives access to the hosted service. Validate the model ID, route and tool configuration with the site's request example before wiring it into your workflow.