← Back to blog
Fable 5.1 Review: Price, Specs, and When to Use It

Fable 5.1 Review: Price, Specs, and When to Use It

Julian Brooks

By Julian Brooks

MyClaw Editorial

MyClaw

Run Best-in-Class AI Agents Now

Run OpenClaw or Hermes Agent on managed MyClaw hosting, so your AI agent stays online, updated, and ready for real work.

AI Takeaway

  • What is Claude Fable 5.1? Anthropic’s current public Mythos-class model. API ID claude-fable-5-1. Shipped 1 September 2026 as the successor to Fable 5.
  • What does it cost? $10 input / $50 output per 1M tokens, same list as Fable 5. Cache reads $0.25 (Fable 5 was $1). Anthropic’s 25% / 45% savings line is versus Fable 5, not Opus 5.
  • Is it the default Claude? No. Official docs: start with Claude Opus 5. Use Fable 5.1 when Opus 5 at higher effort still falls short.
  • Should you switch from Fable 5? Yes for new Fable-class work. Same sticker, cheaper cache, newer checkpoint. Three breaking API changes. Do not blind-swap.

Fable 5.1 at a Glance

Claude Fable 5.1 is the current public Mythos-class checkpoint, not a cheaper SKU and not the default Claude. Official Anthropic pages, checked 3 September 2026:

DetailClaude Fable 5.1
Release date1 September 2026
API model IDclaude-fable-5-1
StatusActive (latest); retirement not sooner than 1 September 2027
Context / max output1M / 128K
Input / outputText and images to text
ThinkingAdaptive, always on; default effort high
Knowledge cutoffJune 2026
API price$10 input / $50 output per 1M tokens
Cache write$12.50 / 5m; $20 / 1h
Cache read$0.25 per 1M
Batch50% off input and output

Claude Mythos 5.1 (claude-mythos-5-1) shares specs and price. It is Project Glasswing only. Public work stays on Fable 5.1. The tokenizer matches Fable 5: the same text produces about 30% more tokens than models older than Claude Opus 4.7.

The window is not new. Fable 5 already held 1M / 128K at $10 / $50. The upgrade is cheaper cache reads, a newer checkpoint, and API behavior that will break a blind swap.

On MyClaw there is no 5.1 model page. The live listing is still Claude Fable 5. Run the IDs the provider actually exposes, and write down any substitution.

What Changed From Fable 5

Introducing Claude Fable 5.1 and Claude Mythos 5.1 \ Anthropic

Same sticker. Cache reads are the discount.

Input and output did not move. If both finish the same job in the same tokens, there is no list-price reason to stay on Fable 5. This is not a same-sticker, higher-bill upgrade. The discount sits in cache.

Cache reads dropped from $1 to $0.25 per million tokens: 0.025 times base input, versus 0.1 on other Claude models. Anthropic estimates that cuts typical token-billed cost by about 25%, and highly agentic workloads by up to about 45%, versus Fable 5. Output is still $50 per million. A chatty loop can erase the cache win.

A cheaper cache hit is still expensive if the ticket does not close. Count:

(model spend + retries + review time + repair time) ÷ accepted tasks

The Fable 5.1 vs Fable 5 page is the upgrade math.

Long-horizon coding, research, and documents

Official docs position Fable 5.1 for demanding reasoning and long-horizon agentic work: multi-hour coding, multistep research, and document, spreadsheet, or slide work that has to stay coherent. The gap versus Fable 5 is widest at higher effort. Anthropic also names vision over dense PDFs, reasoning across the full 1M window, and more reliable computer use. Multilingual performance is on par with Fable 5.

That is a job-shape claim, not a reason to put 5.1 on every autocomplete. Start with Claude Opus 5. Move up when Opus 5 at higher effort still falls short. The Fable 5.1 vs Opus 5 page is that split.

Three breaking API changes

If you already call Fable 5, three changes return HTTP 400 or drop state:

  • Forced tool use is gone. tool_choice set to any or a named tool returns HTTP 400. Keep auto or none. For schema-valid JSON, use strict tool use with auto, or structured outputs.
  • Thinking blocks are one-way. Fable 5.1 can read earlier Claude thinking. Earlier models cannot read 5.1’s. A router that falls back mid-conversation drops those blocks. Dropped blocks are not billed.
  • Editing earlier turns can invalidate thinking. Changing system, tools, or a prior message before a 5.1 thinking block can 400, or drop the block if you opt in. Treat the history as append-only.

Beta extras exist: per-message effort without busting the prompt cache, turn-scoped system messages, and thinking.display: "updates" for progress text between tool calls. They are not required for the ID swap. Do not rebuild the harness around a beta header on day one.

Fable 5.1 also calls tools one-at-a-time more often than Fable 5, writes fewer progress lines at high effort, and is likelier to rewrite a whole file for a small edit. Those are behavior diffs, not scoreboard rows.

Skip the Setup, Run Best-in-Class AI Agents Now

Launch a managed OpenClaw or Hermes Agent workspace in minutes, with always-on hosting, updates, and support handled by MyClaw.

Get Started

Fable 5.1 Pricing: The Bill Still Lives in Output

The rates in the table above are the list. Cache writes did not move. Batch is $5 / $25. Regional and contract rates can differ.

Opus 5 is $5 / $25. Half the sticker. Opus 5 cache reads are $0.50, so a cache-heavy Fable loop can look less than 2× on paper. Finished-job cost still usually favors Opus 5. Do not invert the default because cache math got nicer inside the Fable line.

Fable 5.1 and Mythos 5.1 carry 30-day data retention and are not available under zero data retention unless Anthropic authorizes it. Both are Covered Models, like Fable 5. Handle stop_reason: "refusal" and configure fallback. Permitted fallback targets for 5.1 are Claude Opus 4.8 and Claude Opus 5. You are not billed for a refusal that arrives before any output.

Who Should Use Fable 5.1

Choose Fable 5.1Keep Opus 5Stay off Fable
Jobs Opus 5 already failed at high effort: multi-hour coding, messy research, docs that must stay coherentDefault coding agents, computer use, repeated professional workEveryday autocomplete, bounded patches, anything that does not need Mythos-class spend
New Fable-class work (do not stay on Fable 5)Cache-light volume where $5 / $25 winsMythos 5.1 unless you are in Project Glasswing

Do not migrate a working Opus 5 harness because a Mythos-class ID shipped. Confirm the ID the provider actually exposes, and write down substitutions.

Run Fable 5.1 on a Real Agent Workflow

The model stops if the laptop sleeps, the terminal dies, or the files live in another chat. OpenClaw is the runtime: channels, tools, memory, and scheduled jobs. MyClaw hosts that runtime so the workspace stays up while you swap Opus 5 and Fable 5.1 IDs.

Step 1: Freeze one job Opus 5 already struggled with

Pick a single outcome you already ran on Opus 5 at high and did not like: a failing test, a messy migration, or a sourced research memo that lost the thread. Write the pass/fail rule before either model starts.

Get Started

Step 2: Run Opus 5, then Fable 5.1 at the same effort

Keep files, tools, permissions, and instructions identical. First pass is Opus 5. Second pass is Fable 5.1 at the same high setting when the ID is actually available; otherwise stay on Opus 5 and record the substitution. Track tokens, retries, leftover cleanup, and whether the ticket closed. A coding-agent workflow makes failure visible because tests and diffs do the scoring.

Step 3: Keep the hosted workspace so you can switch

Leave logs, diffs, notes, and artifacts in the same hosted workspace. The next run should start from the work, not a blank chat. MyClaw pricing covers the always-on host. Model tokens still bill separately. Use claude-fable-5-1 unless you already have Mythos 5.1 access.

Final Verdict: Escalate to 5.1. Do Not Default It

Claude Fable 5.1 is the right Fable, not the right default. Official docs still start most work on Opus 5. 5.1 exists for tickets that already failed at high effort: multi-hour coding, messy research, and documents that have to stay coherent. That is a narrower job than the Mythos-class label suggests.

If you are already on Fable 5, move to claude-fable-5-1. Same sticker, cheaper cache, newer checkpoint. Anthropic’s 25% / 45% line is that swap, not a discount versus Opus 5. Output is still the expensive token. Do not quote the cache cut as a reason to skip Opus.

Pay the Mythos-class bill only when a matched run says so. Pin effort, count accepted tasks, and keep a cheaper Claude as fallback. Change the model. Leave the workspace alone.

Skip the Setup, Run Best-in-Class AI Agents Now

Launch a managed OpenClaw or Hermes Agent workspace in minutes, with always-on hosting, updates, and support handled by MyClaw.

Fable 5.1 Review: Price, Specs, and When to Use It | MyClaw.ai