Claude Fable 5 Review: 95% SWE-bench & 1M Tokens (2026)
Claude Fable 5: Anthropic's first Mythos-class AI (June 2026). 1M token context, 95% SWE-bench Verified. At $10/$50 per MTok, tops frontier coding and research.
Claude Fable 5 is Anthropic's first generally available Mythos-class model, released June 9, 2026, with a 1M token context window and 95.0% SWE-bench Verified (the highest of any GA model at launch). Priced at $10/$50 per 1M input/output tokens, it ships with adaptive thinking always on, 128K max output, and day-one availability across Anthropic API, Amazon Bedrock, Vertex AI, and Microsoft Foundry.
Claude Fable 5, released June 9, 2026 by Anthropic, is the first generally available Mythos-class AI with a 1M token context window and 95.0% SWE-bench Verified score. Priced at $10 input and $50 output per 1M tokens, it leads all GA frontier models on coding benchmarks: 80.3% SWE-bench Pro versus Opus 4.8 at 69.2% and GPT-5.5 at 58.6%, with vision, adaptive thinking, and multi-platform availability on Bedrock and Vertex AI.
Provider: Anthropic · Family: Claude Fable
Context window: 1,000,000 tokens · Max output: 128,000
Input modalities: text, image, pdf, tool-calls · Output: text, tool-calls
About Claude Fable 5
Claude Fable 5 is Anthropic's first generally available Mythos-class AI model, released on June 9, 2026. It sits above the Opus tier in Anthropic's model lineup, designed for the most demanding reasoning tasks, autonomous software engineering, knowledge work, vision analysis, and long-horizon agentic execution. Fable 5 is the publicly accessible configuration of a shared Mythos-class architecture: the same underlying weights power Claude Mythos 5, available only to vetted partners through Project Glasswing. Anthropic has not disclosed the parameter count or detailed architecture type. Adaptive thinking is always enabled, meaning the model performs internal reasoning before every response, with depth controlled via an effort parameter rather than a discrete thinking toggle. On software engineering, Fable 5 scores 95.0% on SWE-bench Verified and 80.3% on SWE-bench Pro, ahead of Claude Opus 4.8 at 69.2%, GPT-5.5 at 58.6%, and Gemini 3.1 Pro at 54.2% on SWE-bench Pro. On FrontierCode Diamond, a harder multi-step coding benchmark, Fable 5 reaches 29.3% versus Opus 4.8 at 13.4% and GPT-5.5 at 5.7%. MMLU Pro sits at 91.5%, and HealthBench Professional reaches 66.0% versus Opus 4.8 at 56.9% and GPT-5.5 at 51.8%. On GDPpdf (visual document reasoning without tools), Fable 5 scores 29.8% versus GPT-5.5 at 24.9% and Opus 4.8 at 22.5%. GDPval-AA preference evaluation gives Fable 5 a score of 1932 versus Opus 4.8 at 1890 and GPT-5.5 at 1769. Anthropic did not publish GPQA Diamond, AIME 2025, or ARC-AGI 2 scores at launch. Claude Fable 5 ships with a 1 million token context window at standard pricing, with up to 128,000 output tokens per request. The entire 1M context is billed at the same per-token rate as shorter requests, with no long-context surcharge. Prompt caching applies across the full context window: 5-minute cache writes cost $12.50 per million tokens, 1-hour writes cost $20 per million, and cache reads cost $1 per million, making repeat-context agentic workflows substantially cheaper. Anthropic reports the model stays focused across millions of tokens in long-running agentic tasks, though independent needle-in-haystack evaluations for the full 1M range were not published at launch. The native memory tool is supported, enabling agents to store and retrieve information across conversation boundaries. Fable 5 accepts text, images, PDFs, and tool calls as inputs, and produces text and tool calls as outputs. Vision is fully live at launch: the model reads charts, analyzes screenshots, extracts figures from PDFs, and has demonstrated rebuilding web applications from screenshots and playing Pokemon FireRed using only visual perception. Audio and video inputs are not available. Adaptive thinking replaces the optional extended-thinking parameter from earlier Claude models: every Fable 5 response includes internal reasoning, with depth configurable via the effort parameter. Raw thinking content is never returned; developers can request summarized thinking blocks by setting display to "summarized". Native capabilities include function calling, parallel tool use, the memory tool, task budgets (beta), context editing for long sessions, and compaction for managing context window consumption. Fable 5 is priced at $10 per million input tokens and $50 per million output tokens, making it the most expensive generally available frontier model at launch and exactly double the price of Claude Opus 4.8. The Batch API halves both rates to $5 input and $25 output per million, with results delivered within 24 hours. Prompt caching pays off after a single cache read on the 5-minute TTL: a 100K-token system prompt costs $1.25 to write and $0.10 to read on each subsequent call. US-only inference routing (inference_geo: "us") adds a 1.1x multiplier across all token categories. A daily agentic coding session consuming 2 million input tokens and 400,000 output tokens costs roughly $40 at standard rates, or $20 via the Batch API. From June 9 to June 22, 2026, Fable 5 is included at no extra cost in Claude Pro, Max, Team, and Enterprise subscriptions; starting June 23, continued use requires usage credits billed at API rates. Claude Fable 5 is generally available on the Anthropic Claude API, Amazon Bedrock, Google Vertex AI, Microsoft Azure via Microsoft Foundry, and Claude Platform on AWS. GitHub Copilot added Fable 5 on its launch day, June 9, 2026. Direct API authentication uses a standard Anthropic API key; Bedrock uses AWS IAM, Vertex AI uses GCP IAM, and Foundry uses Azure Active Directory. All Fable 5 sessions carry a mandatory 30-day data retention period: Fable 5 is a Covered Model and zero-data-retention is not available on any platform. Anthropic commits to not training on retained session data, retaining it solely for detecting complex jailbreak attempts and cross-session attack patterns, then deleting it after 30 days. Three classifier systems run on every Fable 5 request before any output is generated. The cybersecurity classifier blocks offensive exploitation, vulnerability discovery, and agentic hacking tasks. The biology and chemistry classifier blocks risky biological research including AAV vector design. The distillation classifier blocks attempts to extract model capabilities for reproduction. When a classifier fires, the API returns stop_reason: "refusal" as an HTTP 200 response with a field identifying which classifier triggered. Developers can pass the fallbacks parameter to retry automatically on Claude Opus 4.8, with fallback credit offsetting the prompt-cache cost of the switch. External bug bounty red-teaming found no universal jailbreaks after over 1,000 hours of effort. Classifiers trigger on fewer than 5% of real-world sessions. Fable 5 is the right choice for teams running autonomous long-horizon software engineering: its 95.0% SWE-bench Verified score leads all generally available models, and Stripe reported compressing months of Ruby codebase migration into a single day. Scientific research applications in drug design, protein hypothesis generation, and genomics benefit from the model's ability to reason across large bodies of literature. Legal and financial teams processing large document corpora benefit from the 1M context window. Teams that should not use Fable 5 include those requiring real-time response, since reasoning overhead adds latency to every request. Audio-first or voice applications have no path because there is no audio modality. Cost-sensitive teams running millions of short queries should use Sonnet 4.6 at $3/$15 per million or Haiku 4.5 at $1/$5 per million instead. Anthropic has not disclosed the training data cutoff date or training data composition for Fable 5. All API sessions carry mandatory 30-day data retention; Anthropic logs all human access to retained data and deletes it after 30 days in most cases. Zero data retention is not available on any platform for Fable 5. The system card published alongside Fable 5 runs 319 pages and covers Firefox zero-day discovery benchmarks, OSS-Fuzz, CyberGym, and CyScenarioBench evaluations. Responsible Scaling Policy (RSP) governs deployment. SOC 2, HIPAA, and GDPR compliance are available through enterprise arrangements on Anthropic-operated and partner cloud platforms. Claude Fable 5 is Anthropic's first generally available Mythos-class deployment, following Claude Mythos Preview (April 2026, restricted access only). Versus Claude Opus 4.8, Fable 5 improves SWE-bench Pro from 69.2% to 80.3%, FrontierCode Diamond from 13.4% to 29.3%, HealthBench Professional from 56.9% to 66.0%, and MMLU Pro to 91.5%. Adaptive thinking replaces the optional extended-thinking toggle: all responses include reasoning with no disable option. New API parameters include effort for reasoning depth control, task-budgets (beta) for token-aware agentic planning, and fallbacks for automatic classifier-triggered model switching. The API model ID is claude-fable-5.
Pricing
$10/$50 per 1M input/output tokens. Batch API: $5/$25 per 1M (50% off, async within 24h). Cache read: $1/MTok (0.1x input rate). 5-min cache write: $12.50/MTok; 1-hr cache write: $20/MTok. US-only inference (inference_geo: 'us') adds 1.1x multiplier on all token categories. Free on Claude Pro/Max/Team/Enterprise plans June 9-22, 2026; requires usage credits from June 23 onward.
Key Features
- Adaptive Thinking Always On: Internal reasoning is applied to every request by default, with depth controlled via the effort parameter. Unlike earlier Claude models, thinking cannot be disabled on Fable 5, and raw chain-of-thought is never returned.
- 1M Token Context Window: Handles up to 1 million tokens in a single context at standard per-token pricing with no long-context surcharge. Supports processing entire large codebases, legal document sets, or research corpora in one API call.
- Safety Classifier Fallback System: Three built-in classifiers (cybersecurity, biology/chemistry, distillation) detect high-risk requests and route them to Claude Opus 4.8 automatically, returning HTTP 200 with stop_reason: 'refusal' and the triggering classifier identified.
- Vision and Document Understanding: Reads images, charts, screenshots, and PDFs natively at launch. Demonstrated rebuilding web application interfaces from screenshots and playing Pokemon FireRed using only visual input with no text.
- Multi-Platform Day-One Availability: Generally available on Anthropic API, Amazon Bedrock, Google Vertex AI, Microsoft Foundry, Claude Platform on AWS, and GitHub Copilot simultaneously on June 9, 2026, with no waitlist for API access.
Pros
- 95.0% SWE-bench Verified and 80.3% SWE-bench Pro, highest of any generally available model at launch, with a 21.7-point margin over GPT-5.5 on the harder Pro split.
- 1M context window at standard pricing with no surcharge, plus adaptive thinking that maintains coherence across full-million-token agentic sessions.
- Day-one availability across five major platforms (Anthropic API, Bedrock, Vertex, Foundry, GitHub Copilot) covering every major enterprise cloud stack without waitlists.
Cons
- Mandatory 30-day data retention with no zero-retention option on any platform, blocking adoption in sensitive sectors with strict data handling requirements.
- At $10/$50 per million tokens, the most expensive generally available frontier model at launch, making it the wrong choice for high-volume short-query workloads.
- No audio input or output; adaptive thinking adds latency to every request regardless of complexity, ruling out real-time voice-first and latency-critical applications.
Benchmarks
- gdppdf: 29.8
- mmlu pro: 91.5
- gdpval aa: 1932
- swe bench pro: 80.3
- swe bench verified: 95
- frontier code diamond: 29.3
- health bench professional: 66
Frequently Asked Questions
What is Claude Fable 5 and who built it?
Claude Fable 5 is Anthropic's first generally available Mythos-class AI model, released on June 9, 2026. It sits above the Opus tier in Anthropic's model lineup and represents a qualitative step beyond Claude Opus 4.8 in reasoning depth, coding capability, and agentic autonomy. The model is built for demanding software engineering, long-horizon autonomous tasks, knowledge work, vision analysis, and scientific research at the frontier. Fable 5 is the general-access configuration of a shared underlying Mythos-class architecture: the same weights power Claude Mythos 5, available only to vetted partners working in cybersecurity and biology research through Project Glasswing. Anthropic has not disclosed the parameter count or detailed architecture type; adaptive thinking is always enabled, making Fable 5 qualitatively different from earlier Claude models where thinking was an optional toggle. The API model ID is claude-fable-5, available on the Anthropic API, Amazon Bedrock, Google Vertex AI, Microsoft Foundry, and GitHub Copilot. In customer deployments at launch, Stripe compressed months of engineering work into one day, and Cognition's FrontierCode benchmark ranked it highest among frontier models at medium effort.
How much does Claude Fable 5 cost per 1M tokens?
Claude Fable 5 costs $10 per million input tokens and $50 per million output tokens, making it the most expensive generally available frontier model at launch and exactly double the price of Claude Opus 4.8 at $5/$25. Cache reads cost $1 per million tokens (0.1x the input rate), and there are two cache write tiers: $12.50 per million for a 5-minute TTL and $20 per million for a 1-hour TTL. The Batch API delivers a 50% discount on both rates: $5 input and $25 output per million, with results returned within 24 hours, making high-volume non-urgent workloads significantly cheaper. For a worked example, a daily coding agent consuming 2 million input tokens and 400,000 output tokens would cost $40 at standard rates or $20 via the Batch API. US-only data routing using inference_geo: 'us' adds a 1.1x multiplier across all token categories including cache operations. From June 9 to June 22, 2026, Fable 5 is included at no extra cost in Claude Pro, Max, Team, and Enterprise plans; from June 23, continued use requires usage credits billed at API rates. Self-hosting is not available: Fable 5 is closed-weights and API-only.
What is Claude Fable 5's context window and max output?
Claude Fable 5 supports a 1 million token context window by default, the largest standard context in Anthropic's publicly released model lineup. The maximum output per request is 128,000 tokens, sufficient for generating large documents, complete codebases, or detailed research reports in a single API call. The entire 1M context window is billed at the standard per-token input rate with no surcharge for long-context requests, unlike some providers who charge a premium for extended context. Anthropic reports the model stays focused across millions of tokens in long-running agentic tasks, though independent third-party needle-in-haystack evaluations across the full 1M range had not been published at launch. Prompt caching applies across the full context window: a 100K-token system prompt cached for 5 minutes costs $1.25 to write and $0.10 to read on each subsequent call, reducing repeat-context costs by 90%. For comparison, Claude Opus 4.8 also supports 1M context, while the full 1M range at standard pricing with no surcharge is a distinguishing feature versus some competing providers. The model supports the memory tool natively, enabling knowledge persistence across session boundaries in long-running agentic applications.
How does Claude Fable 5 compare on benchmarks vs GPT-5.5?
On SWE-bench Pro, Claude Fable 5 scores 80.3% versus GPT-5.5 at 58.6%, a gap of 21.7 percentage points that translates to completing substantially more real-world engineering tasks autonomously without human intervention. On FrontierCode Diamond, Fable 5 reaches 29.3% versus GPT-5.5 at 5.7%, meaning Fable 5 solves roughly five times more hard multi-step coding challenges at medium effort. On HealthBench Professional (medical knowledge evaluation), Fable 5 leads at 66.0% versus GPT-5.5 at 51.8%. On GDPpdf (visual document reasoning without tools), Fable 5 scores 29.8% against GPT-5.5's 24.9%. GDPval-AA preference evaluation gives Fable 5 a score of 1932 versus 1769 for GPT-5.5, indicating stronger human preference in head-to-head comparisons. Anthropic did not publish GPQA Diamond, AIME 2025, or ARC-AGI 2 scores for Fable 5 at launch, so direct comparison on those benchmarks is not available from official data. GPT-5.5 supports audio input, which Fable 5 does not, giving it an advantage for voice-first applications. The benchmark advantage on coding and knowledge work is vendor-reported; independent verification was in progress as of launch.
Is Claude Fable 5 open source or proprietary?
Claude Fable 5 is fully proprietary: the weights are not released, and access is available only through API endpoints operated by Anthropic and its cloud partners. There is no option to download or self-host the model, run it in an air-gapped environment, or fine-tune the base weights. Fable 5 is accessible via the Anthropic Claude API (api.anthropic.com), Amazon Bedrock (billed through AWS), Google Vertex AI (billed through GCP), Microsoft Azure via Microsoft Foundry, Claude Platform on AWS, and GitHub Copilot. Authentication methods differ by platform: Anthropic API uses API keys, Bedrock uses AWS IAM roles, Vertex AI uses GCP service accounts, and Foundry uses Azure Active Directory. The license for API usage is governed by Anthropic's commercial terms; no open-source or research license exists for this model. For teams that need on-device inference or air-gapped deployment, there is no path with Fable 5. Open-weights alternatives at somewhat lower capability include Llama 4 and Qwen3-Max, though neither matched Fable 5's SWE-bench Pro scores at the time of this writing.
What modalities does Claude Fable 5 support?
Claude Fable 5 accepts text, images, PDFs, and tool calls as inputs. Output modalities are text and tool calls. Vision is fully live at launch: the model reads charts, interprets diagrams, analyzes product screenshots, extracts data from PDF figures, and has demonstrated rebuilding web app interfaces from uploaded images and playing Pokemon FireRed using only screenshot input. Audio input and output are not supported on Fable 5 or Mythos 5. Video input is not supported. Function calling is native and supports parallel tool execution across multiple tools in a single turn. Structured JSON output is supported. The model natively supports the memory tool for persistent cross-session storage, task budgets (beta) for token-aware agentic planning, and context editing for pruning tool results in long agentic loops. Computer use (screen-reading and click automation via the computer use beta) is not listed as a supported feature in Fable 5 launch documentation. Compared to GPT-5.5, which supports audio input, Fable 5 is disadvantaged for voice-first applications but leads on agentic coding and long-document tasks.
Does Claude Fable 5 train on user data?
Claude Fable 5 does not train on API inputs. However, all Fable 5 sessions are subject to mandatory 30-day data retention: the API retains inputs and outputs for safety monitoring, and this cannot be opted out of regardless of plan or contract. Anthropic commits explicitly to not using retained Fable 5 data to train new Claude models or for any purpose other than safety monitoring (detecting complex jailbreaks and cross-session attack patterns). All human access to retained data is logged, and data is deleted after 30 days in most cases. Unlike earlier Claude models that offered zero-data-retention for enterprise customers, zero-data-retention is not available for Fable 5 or Mythos 5 on any platform including Bedrock and Vertex AI. This is a significant constraint for teams in legal, defense, or healthcare sectors with strict data handling requirements. SOC 2 Type II, HIPAA eligibility, and GDPR compliance are available through enterprise arrangements on Anthropic-operated platforms; partner-operated platforms (Bedrock, Vertex) provide their own compliance postures. The EU AI Act classifies Fable 5 under general-purpose AI provisions with systemic risk obligations.
Who is Claude Fable 5 best for and who should avoid it?
Claude Fable 5 is best for engineering teams running long-horizon autonomous coding tasks: at 95.0% SWE-bench Verified and 80.3% SWE-bench Pro, it outperforms every other generally available model on the primary agentic coding benchmark. Scientific research organizations benefit from its ability to generate novel protein hypotheses, design drug candidates, and reason across large genomics datasets, with Anthropic reporting a 10x acceleration in aspects of drug design in early deployments. Legal and financial teams processing large document corpora gain from the 1M context window at standard pricing. Teams that should use alternatives include those running real-time voice applications, which must use a separate model since Fable 5 has no audio modality. Cost-sensitive teams running millions of short daily queries will find Claude Opus 4.8 at $5/$25 per million or Sonnet 4.6 at $3/$15 a much better fit. Teams in sectors requiring zero data retention (certain defense, legal, healthcare contexts) cannot use Fable 5 since mandatory 30-day retention cannot be disabled on any platform. Applications where latency matters more than reasoning quality should consider Sonnet 4.6 or Haiku 4.5, which do not add reasoning overhead to every request.