What is Claude Fable 5?
Claude Fable 5 is Anthropic's June 2026 model for long-running reasoning, coding, vision, and document work. It is the first model in the Claude 5 family and opens a new Mythos class that sits above the Opus line. It shares its core model with Claude Mythos 5, but adds stricter safeguards for sensitive cyber, biology, chemistry, and distillation requests; Mythos 5 stays limited to approved organizations. Anthropic restored access on July 1, 2026 after a short pause in June. On Unifically it runs as anthropic/claude-fable-5 with a 1M token context window, up to 128k output tokens, and text plus image input.
Key features of Claude Fable 5
Top rank on Arena Text
Claude Fable 5 ranks #1 on Arena's Text board with a 1509±9 score, ahead of Claude Opus 4.6 Thinking and Claude Opus 4.7 Thinking. That makes it a strong pick for open-ended reasoning, coding, writing, and analysis.
Top rank on Arena Vision
It also ranks #1 on Arena's Vision board at 1311±14. Use it when screenshots, charts, diagrams, and visual evidence are part of the task, not an afterthought.
1M token context, 128k output
The working window holds a large repo, a long document set, or a full agent transcript at once, and a single response can run up to 128k tokens.
Always-on adaptive thinking
Adaptive thinking is the only thinking mode. The model decides how much to reason per request; raw chain of thought is not returned, only a readable summary when thinking blocks appear.
Built for long-horizon agent work
Fable 5 is tuned for tasks that run across many steps: planning, tool calls, memory, and progress checks without losing state mid-run.
Best for
Large codebase migrations
Repo-wide changes and multi-step code review where the model must hold the whole picture.
Long document analysis
PDFs, charts, tables, and reports read together and returned as one structured brief.
Agent workflows
Planning, tools, memory, and progress checks across runs that would restart a smaller model.
Vision-heavy reasoning
Screenshots, diagrams, and scientific figures as first-class input next to the prompt.
Deep research briefs
Synthesis across many sources held in view at once inside the 1M window.
High-stakes enterprise work
Tasks where the model must check its own output before it lands.
Use cases
Build a coding agent that inspects a repo, plans a migration, edits files, writes tests, and reviews its own work in one session. Use it as the reading layer in a document intelligence product that takes long filings, contracts, spreadsheets, and charts and returns a structured brief. It also fits research copilots that keep a large body of source material in view while drafting. For product teams, Fable 5 works well behind bug triage, PR review, technical planning, and multi-step operations where a smaller model loses track.
Limitations
Fable 5 carries safeguards that can decline or redirect sensitive requests in cybersecurity, biology, chemistry, and distillation. They are tuned conservatively, so some harmless requests in those areas can get caught.
It also requires 30-day data retention for safety monitoring and is not available under zero data retention. For workloads with strict retention rules, that matters more than raw model quality.
Claude Fable 5 vs Claude Opus 4.8
Both list a 1M context window and 128k max output, but Fable 5 ranks higher on Arena's Text, Vision, Document, and Search boards. On Text, Fable 5 sits at #1 with 1509±9 while Claude Opus 4.8 ranks #13 at 1477±6. On Vision, Fable 5 is #1 at 1311±14 against #13 at 1280±9 for Opus 4.8. Use Fable 5 when quality matters more than speed; Opus 4.8 remains the faster, cheaper option for day-to-day work.
When to use Claude Fable 5
Use Claude Fable 5 for the hard work you would otherwise split into many smaller prompts. It is at its best when the model must read a lot, reason across many steps, call tools, inspect visual input, and return a finished result with fewer restarts. For simple chat, extraction, and short drafts, a smaller model is usually enough.
API examples
Call Claude Fable 5 from any language by POSTing to /v1/chat/completions, the OpenAI-compatible endpoint shared by every language model on the platform. Full parameter docs live at docs.unifically.com/models/llm/anthropic/claude-fable-5.
curl -X POST https://api.unifically.com/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "anthropic/claude-fable-5",
"messages": [
{ "role": "user", "content": "Read this migration plan and list the risks in order of severity." }
]
}'
The response comes back synchronously with the completion. Set "stream": true to receive tokens as they generate.
FAQs
People also ask
anthropic/claude-fable-5, called through the OpenAI-compatible POST /v1/chat/completions endpoint with your Unifically API key.
A 1M token context window, with up to 128k output tokens per response. That covers a large repo, a long filing pack, or a full agent session in one call.
Yes. It takes text and image input in the same request, so screenshots, charts, diagrams, and scanned documents can sit next to the prompt. Output is text.
It is strong at both. It ranks
They share the same underlying model. Fable 5 is the generally available version and includes safety classifiers for sensitive domains; Mythos 5 is limited to approved organizations.
No. Adaptive thinking is the only thinking mode, and thinking blocks come back as a readable summary or are omitted. Raw chain of thought is not returned.
Two things. Safeguards in cyber, biology, chemistry, and distillation areas can decline or redirect requests, and the model requires 30-day data retention, so it is not available under zero-data-retention terms.
