Claude Fable 5 Review: Price, Benchmarks & Claude Code Guide (2026)
Stripe ran a 50-million-line Ruby codebase migration in one day. Without Claude Fable 5, that task would have taken a full engineering team over two months. That single data point tells you everything you need to know about what just shipped.
On June 9, 2026, Anthropic launched Claude Fable 5, the first publicly available model in its new Mythos class. This is the same underlying architecture as Claude Mythos 5, the model Anthropic held back from the public for months over cybersecurity concerns. Fable 5 is Mythos made accessible: wrapped in safety classifiers, priced at $10/$50 per million tokens, and available today on the API, Claude Code, Amazon Bedrock, and GitHub Copilot.
I've gone through everything Anthropic published, tested the API model ID, and tracked down the real numbers. This is the complete guide, with no filler.
What Is Claude Fable 5?
Claude Fable 5 is Anthropic's most capable publicly available AI model, launched June 9, 2026. It belongs to the new Mythos class, a model tier above the Opus class that Anthropic introduced for the first time with this release. Fable 5 is state-of-the-art across software engineering, knowledge work, vision, and scientific research, and its lead over previous models grows the longer and more complex the task gets.
The model's API identifier is claude-fable-5. It supports a 1 million token context window and up to 128k output tokens per request. For any task that used to require splitting context or losing thread, Fable 5 handles it in a single run.
Key specs: Model ID: claude-fable-5 | Context: 1M tokens | Output: 128k tokens | Launch: June 9, 2026
The name "Fable" refers to the generally available version of Mythos. Think of it as Mythos with guardrails enabled for public use. Anthropic says the two models share the same underlying weights; the difference is access control and safety classifier behavior.
How Claude Fable 5 Differs from Claude Mythos 5
Claude Fable 5 and Claude Mythos 5 are the same model underneath. The practical difference: Fable 5 ships with safety classifiers that watch every session and reroute high-risk queries to Claude Opus 4.8 instead of answering directly. Mythos 5 has those classifiers lifted in specific areas.
According to Anthropic, the classifiers trigger in fewer than 5% of sessions on average. When they do fire, the user gets a response from Opus 4.8 (still a strong model) and a notification that their query was handled by a different model.
Restricted topics for Fable 5 include cybersecurity attack vectors, synthesis routes for controlled biological agents, certain chemistry topics, and model distillation requests. Everything else, including defensive security research, coding, document analysis, and knowledge work, runs on full Mythos-class capability.
Mythos 5 is restricted to organizations approved under Project Glasswing, Anthropic's initiative for infrastructure defenders and government-cleared organizations. If you're not in that program, Fable 5 is your ceiling for now, and it's a very high ceiling.
Claude Fable 5 Benchmarks
The numbers are real and they're significant. Fable 5 scores 80.3% on SWE-Bench Pro, the agentic coding benchmark Anthropic considers the most meaningful for production software work. For comparison, Claude Opus 4.8 scores 69.2% on the same benchmark, and GPT-5.5 scores 58.6%.
That 21.7-point gap between Fable 5 and GPT-5.5 on SWE-Bench Pro represents hundreds of real software engineering tasks where Fable 5 ships working code and GPT-5.5 doesn't.
Key Benchmark Numbers (Fable 5 vs Opus 4.8)
- SWE-Bench Pro (agentic coding): 80.3% vs 69.2%�/u��u�GDPval-AA (knowledge work Elo): 1932 vs 1890�/u��u�OSWorld-Verified (computer use): competitive with Opus 4.8�/u��u�Legal Agent Benchmark: 13.3% vs 10.4% (GPT-5.5: 2.1%)�/u��u�FrontierCode Diamond (token efficiency): 29.3% at high effort�/u��u�Finance Benchmark (Hebbia): highest score of any model tested
The token efficiency numbers matter for cost. On Cognition's FrontierCode evaluation, Fable 5 scores highest among frontier models even at medium effort, which means it reaches the correct answer using fewer tokens than competing models. At the $10/$50 pricing, that partially offsets the premium over Opus 4.8.
Vision is another area where Fable 5 leads. It can rebuild a web app from screenshots alone, extract precise numbers from detailed scientific figures, and completed Pokemon FireRed using only raw game screenshots with no navigation aids or helper harness. Previous Claude models needed significant scaffolding to do the same thing.
Claude Fable 5 Price and API Access
Claude Fable 5 costs $10 per million input tokens and $50 per million output tokens. That is exactly double Claude Opus 4.8's standard rate of $5/$25. Batch pricing is $5/$25 per million tokens, and prompt caching brings a 90% discount on cached input tokens.
The model is available today on the Claude API using the model string claude-fable-5. It is also available on Amazon Bedrock (US East N. Virginia and Europe Stockholm regions), Vertex AI, Microsoft Foundry, and GitHub Copilot (Pro+, Max, Business, Enterprise).
API Quick Start
To call Fable 5 from the Claude API:
- Model string: claude-fable-5�/u��u�Max tokens: up to 128000�/u��u�Context window: 1,000,000 tokens�/u��u�Pricing: $10 input / $50 output per million tokens
Data retention is a required constraint. Fable 5 is a covered model with 30-day data retention, which Anthropic needs to operate the safety classifiers. Zero data retention is not available for this model. If your workload requires ZDR, use Claude Opus 4.8.
Don't just use ChatGPT. Learn to build custom LLM agents, RAG pipelines, and full-stack Generative AI apps in our intensive 8-week program.
Is Claude Fable 5 Free?
Partially, and only for a limited window. Anthropic included Fable 5 in Claude Pro, Max, Team, and seat-based Enterprise plans at no extra cost from June 9 through June 22, 2026. Starting June 23, continued use requires usage credits, which bill at standard API rates once plan limits are exhausted.
The free tier on claude.ai does not include Fable 5. Anthropic says it intends to restore Fable 5 as a standard subscription feature as capacity allows, but has not committed to a date.
I'd treat the free window as a trial, not as the long-term pricing reality. If you want to pressure-test Fable 5 on your real workloads before deciding whether the $10/$50 API rate makes sense for your use case, June 9-22 is your window.
How to Use Claude Fable 5 in Claude Code
Claude Fable 5 is available in Claude Code. Anthropic made Fable 5 accessible across all of its developer surfaces on launch day, including Claude Code CLI and Claude Code for web.
Selecting Fable 5 in Claude Code CLI
To use Claude Fable 5 in the Claude Code CLI, pass the model flag when starting a session:
claude --model claude-fable-5You can also set it as your default model in your Claude Code configuration file so you don't need to pass it on every invocation.
Claude Code for Web
In Claude Code for the web, open the model picker and select Claude Fable 5 from the dropdown. The model appears in the list once it's enabled for your workspace. If you're on a Team or Enterprise plan, your admin may need to enable Fable 5 in workspace settings.
GitHub Copilot
Fable 5 is rolling out to GitHub Copilot Pro+, Max, Business, and Enterprise users across VS Code (all modes: chat, ask, edit, agent), Visual Studio, JetBrains, Xcode, Eclipse, the CLI, github.com, and GitHub Mobile. Business and Enterprise admins must enable the Claude Fable 5 policy in Copilot settings, as it is off by default.
One important note for GitHub Copilot: Fable 5 requires data retention to operate the safety classifiers. Anthropic retains prompts and outputs for up to 30 days. The retained data is not used to train models, but if this is a concern for your organization, Opus 4.8 remains available as an alternative.
In GitHub's internal testing, Fable 5 completed equivalent autonomous coding work with fewer tool calls and lower token consumption than previous Opus-tier models. For long-horizon coding tasks, that matters.
Claude Fable 5 vs Claude Opus 4.8: Which Should You Use?
Claude Opus 4.8 is still an excellent model and the right choice for a large portion of workloads. The honest comparison:
- Use Fable 5 for: long-horizon autonomous coding, large codebase migrations, complex document analysis spanning hundreds of pages, finance and legal reasoning requiring senior-level judgment, detailed vision tasks, and scientific research workflows where you need hours of autonomous work without intervention.�/u��u�Use Opus 4.8 for: cost-sensitive production workloads where Fable 5's pricing would meaningfully affect your unit economics, any task that requires zero data retention, high-volume API calls where the 2x price difference adds up, and routine coding or chat where Fable 5's extra capability isn't the bottleneck.
The Stripe example is instructive here. A two-month engineering task compressed to one day is not a marginal improvement. It's a different category of output. For tasks at that complexity level, Fable 5 is the right model regardless of price. For tasks that Opus 4.8 already handles reliably, the extra cost doesn't buy you much.
Fable 5's lead over Opus 4.8 also grows with task complexity. On simpler benchmarks, the gap is smaller. On the hardest agentic tasks, Fable 5 pulls away significantly. This is a model for hard problems.
Is Claude Better Than ChatGPT in 2026?
On software engineering, Claude Fable 5 has a significant lead over GPT-5.5. An 80.3% vs 58.6% score on SWE-Bench Pro is a 21.7-point gap that reflects hundreds of real coding tasks where one model succeeds and the other doesn't.
On the Legal Agent Benchmark, Fable 5 scores 13.3% against GPT-5.5's 2.1%. These are intentionally hard tasks designed to be barely solvable, and Fable 5 is more than six times further along than GPT-5.5 on that benchmark.
For knowledge work, Fable 5 holds the top position on Hebbia's Finance Benchmark for senior-level reasoning. IMC reported that Fable 5 "aced their trading-analysis evaluations nearly across the board," including factual lookup, conceptual reasoning, root-cause analysis, and expected-value analysis.
GPT-5.5 remains competitive on certain tasks, particularly those where it has been optimized for specific formats or instruction-following patterns. But for complex, long-horizon agentic work, which is the direction the entire industry is moving, the current evidence puts Fable 5 ahead.
Safeguards: What Gets Blocked and What Doesn't
Fable 5 is Mythos with classifiers enabled. When the classifier detects a high-risk query, the model routes to Opus 4.8 instead and notifies the user. Anthropic says this fallback triggers in fewer than 5% of sessions on average, and they've tuned the classifiers conservatively to minimize false positives on legitimate requests.
Topics that trigger the classifier include offensive cybersecurity (attack tools, exploit development), synthesis routes for controlled biological or chemical agents, and model distillation requests. Defensive security research, code review, pen testing conceptually, and standard development work all run on full Fable 5 capability.
Anthropic is open about the tradeoff: they've set the classifiers conservatively, which means some harmless requests will get routed to Opus 4.8. They're working to reduce false positives as they learn how Fable 5 is actually being used in the field.
Recommended Reads
If you found this useful, these posts from Build Fast with AI go deeper on related topics:
- Claude Mythos 5 Review: Anthropic's Most Powerful Model (2026)
- Claude Managed Agents Dreaming Explained
- Claude Design: Complete Guide for Non-Designers (2026)
- Best AI Models of May 2026: Full Leaderboard and Rankings
- Claude AI 2026: Models, Features, Desktop and More
Want to build AI agents that use models like Claude Fable 5 in production?
Join Build Fast with AI's Gen AI Launchpad — an 8-week structured program to go from 0 to 1 in Generative AI.
Frequently Asked Questions
What is Fable 5 in Claude?
Claude Fable 5 is Anthropic's first publicly available Mythos-class model, launched June 9, 2026. It shares the same underlying architecture as Claude Mythos 5 but ships with safety classifiers that redirect high-risk queries to Opus 4.8. The API model ID is claude-fable-5. It supports a 1M token context window and 128k output tokens.
What does Claude Fable 5 do?
Fable 5 excels at long-horizon autonomous coding, large codebase migrations, complex document analysis, vision tasks including rebuilding web apps from screenshots, senior-level financial and legal reasoning, and scientific research. Its lead over previous models grows with task complexity. Stripe reported that Fable 5 compressed a two-month engineering task into a single day.
Is Claude Fable 5 free?
Fable 5 is included at no extra cost on Claude Pro, Max, Team, and seat-based Enterprise plans from June 9 through June 22, 2026 only. After June 23, usage requires usage credits. The claude.ai free tier does not include Fable 5. API access is paid from day one at $10/$50 per million tokens.
What is the Claude Fable 5 price?
Fable 5 costs $10 per million input tokens and $50 per million output tokens on the Claude API. Batch pricing is $5/$25 per million tokens. Prompt caching brings a 90% discount on cached input tokens. This is double the standard Opus 4.8 rate of $5/$25.
How do I use Claude Fable 5 in Claude Code?
In the Claude Code CLI, use: claude --model claude-fable-5. In Claude Code for web, select Claude Fable 5 from the model picker. GitHub Copilot users (Pro+, Max, Business, Enterprise) can select it from the model picker in VS Code, Visual Studio, JetBrains, and other supported surfaces. Business and Enterprise admins must enable the policy first.
What is Claude Mythos 5 and how is it different from Fable 5?
Claude Mythos 5 and Fable 5 use the same underlying model. The difference is that Mythos 5 has the safety classifiers lifted in specific areas, giving it unrestricted access to cybersecurity capabilities that Fable 5 blocks. Mythos 5 is restricted to organizations approved through Project Glasswing and is not generally available.
Is Claude Fable 5 available on Claude Code?
Yes. Anthropic made Fable 5 available on Claude Code CLI, Claude Code for web, and Claude Cowork on launch day. You can select it via the model picker or pass the --model claude-fable-5 flag in the CLI. Free access on subscription plans runs through June 22, 2026.
Is Claude better than ChatGPT?
For complex agentic coding tasks, Claude Fable 5 scores significantly higher than GPT-5.5: 80.3% vs 58.6% on SWE-Bench Pro. On the Legal Agent Benchmark, Fable 5 scores 13.3% vs GPT-5.5's 2.1%. Both models are strong, but for long-horizon autonomous work, the current benchmarks favor Fable 5.
References
- Claude Fable 5 and Claude Mythos 5 - Anthropic Official Announcement - Anthropic, June 9, 2026�/o��o�Introducing Claude Fable 5 and Claude Mythos 5 - Claude API Docs - Anthropic Platform Docs�/o��o�Anthropic's Claude Fable 5 is a version of Mythos the public can access today - TechCrunch, June 9, 2026�/o��o�Anthropic Claude Fable 5 on AWS - AWS Blog, June 9, 2026�/o��o�Claude Fable 5 is generally available for GitHub Copilot - GitHub Changelog, June 9, 2026�/o��o�Claude Fable 5 Benchmarks and Prompting Guide - Agentpedia, June 9, 2026�/o��o�Anthropic releases Mythos-like AI model to the public, Claude Fable 5 - CNBC, June 9, 2026
