# Claude Fable 5 ![Organization](https://img.shields.io/badge/Org-Anthropic-d97757) ![License](https://img.shields.io/badge/License-Proprietary-d29922) ![Modality](https://img.shields.io/badge/Modality-Text%20%2B%20Vision-8957e5) ![Status](https://img.shields.io/badge/Access-Suspended-red) > **Quick answer:** Claude Fable 5 is a specialised model from Anthropic's Fable line, distinct from the standard Haiku/Sonnet/Opus tiers. It achieves 95% on SWE-bench Verified, 88% on BrowseComp, and 40% on AA Omniscience. As of June 2026, access is suspended under a US government export control directive. Teams requiring similar capabilities should evaluate Claude Opus 4 or Claude Sonnet 4. ## Access Status > **Important:** As of June 2026, access to Claude Fable 5 is suspended under a US government export control directive. See [Anthropic's statement](https://www.anthropic.com/news/fable-mythos-access) for current status. Benchmark data below is preserved for reference. ## At a Glance **Where Claude Fable 5 leads** - 95% SWE-bench Verified — among the highest scores on the software engineering benchmark - 88.0% BrowseComp — strong agentic web research and retrieval - 84.2% MMMU-Pro — high multimodal understanding performance - 83.3% MCP Atlas — strong tool-use and API orchestration - 40% AA Omniscience — leading knowledge retrieval score **Where it lags** - Access currently suspended under US export control directive - 48.7% Agent's Last Exam — below the top frontier models (GPT-5.6 Sol: 52.7%) - Pricing not publicly disclosed - Limited deployment context compared to the standard Claude series **Best for:** Reference benchmarking; teams with prior access continuing existing deployments. ## What Claude Fable 5 Is Claude Fable 5 is part of Anthropic's expanded model family beyond the standard Haiku/Sonnet/Opus tiers. Anthropic has developed specialised model lines — including Fable and Mythos — targeting specific capability profiles or deployment contexts that differ from the general-purpose commercial lineup. Fable 5's benchmark profile positions it as a highly capable model for software engineering and agentic tool-use tasks. Its 95% SWE-bench Verified score is among the highest publicly recorded and suggests particular strength in multi-step code modification and test-driven development workflows. The 88% BrowseComp score similarly indicates strong capability for web-based research and retrieval tasks. In June 2026, Anthropic issued a statement noting that the US government issued an export control directive suspending access to both Fable 5 and Mythos 5. This limits general availability indefinitely pending resolution of the directive. ## Specifications | Field | Value | |---|---| | Organization | Anthropic | | Parameters | Undisclosed | | Context window | Undisclosed | | Architecture | Undisclosed (Fable line) | | License | Proprietary (API only) | | Modality | Text + Vision (multimodal) | | Knowledge cutoff | October 2025 (estimated) | | Access status | Suspended (US export control, June 2026) | ## Pricing Pricing for Claude Fable 5 is not publicly disclosed. Access is currently suspended — see Anthropic's official statement for current status. ## Public Benchmark Scores | Benchmark | Score | Source | Date | |---|---|---|---| | [SWE-bench Verified](/benchmarks/swe-bench-verified) | 95.0% | Benchgen evaluation | 2026-06 | | [BrowseComp](/benchmarks/browsecomp) | 88.0% | Benchgen evaluation | 2026-06 | | [MMMU-Pro](/benchmarks/mmmu-pro) | 84.2% | Benchgen evaluation | 2026-06 | | [MCP Atlas](/benchmarks/mcp-atlas) | 83.3% | Benchgen evaluation | 2026-06 | | [Global MMLU Lite](/benchmarks/global-mmlu-lite) | 93.3% | Benchgen evaluation | 2026-06 | | [AA Omniscience](/benchmarks/aa-omniscience) | 40% | Benchgen evaluation | 2026-06 | | [Agent's Last Exam](/benchmarks/agents-last-exam) | 48.7% | Benchgen evaluation | 2026-06 | | [SimpleQA](/benchmarks/simpleqa) | 68.3% | Benchgen evaluation | 2026-06 | | [IFBench](/benchmarks/ifbench) | 63.5% | Benchgen evaluation | 2026-06 | | [LHTB](/benchmarks/lhtb) | 48.7% | Benchgen evaluation | 2026-06 | *Scores from Benchgen evaluations conducted prior to the access suspension. Shown for reference.* ## Claude Fable 5 vs Alternatives | Model | SWE-bench Verified | BrowseComp | Agent's Last Exam | Access | |---|---|---|---|---| | **Claude Fable 5** | 95.0% | 88.0% | 48.7% | Suspended | | [GPT-5.6 Sol](/models/openai/gpt-5-6-sol) | — | 90.84% | 52.7% | Available | | [Claude Opus 4.6](/models/anthropic/claude-opus-4-6) | — | — | — | Available | | [Claude Sonnet 4](/models/anthropic/claude-sonnet-4) | — | — | — | Available | Claude Fable 5's 95% SWE-bench Verified score is its most notable result — among the highest published. For teams unable to access Fable 5, Claude Opus 4.6 is Anthropic's recommended alternative for high-capability coding and agentic tasks. ## Related Models & Benchmarks - Compare: [Claude Opus 4.6](/models/anthropic/claude-opus-4-6) · [Claude Sonnet 4.6](/models/anthropic/claude-sonnet-4-6) · [GPT-5.6 Sol](/models/openai/gpt-5-6-sol) - Benchmarks: [SWE-bench Verified](/benchmarks/swe-bench-verified) · [BrowseComp](/benchmarks/browsecomp) · [Agent's Last Exam](/benchmarks/agents-last-exam) · [MCP Atlas](/benchmarks/mcp-atlas) ## Frequently Asked Questions

What is Claude Fable 5?

Claude Fable 5 is a specialised model from Anthropic's Fable line, distinct from the standard Haiku/Sonnet/Opus tiers. It achieves 95% on SWE-bench Verified and 88% on BrowseComp. Access is currently suspended under a US government export control directive as of June 2026.

Can I access Claude Fable 5?

No. As of June 2026, access to Claude Fable 5 is suspended under a US government export control directive. See Anthropic's official statement for the most current information on access status.

What is Claude Fable 5's SWE-bench Verified score?

Claude Fable 5 scored 95% on SWE-bench Verified — among the highest recorded scores on this software engineering benchmark.

What is an alternative to Claude Fable 5?

Anthropic recommends Claude Opus 4 or Claude Sonnet 4 as alternatives for teams that previously used Fable 5. GPT-5.6 Sol is a strong alternative for software engineering and agentic tasks.

Is Claude Fable 5 open source?

No. Claude Fable 5 is a proprietary, API-only model. There are no open weights.

--- *Specs and scores sourced from Anthropic's official communications and Benchgen evaluations; third-party benchmark scores attributed inline. Access status as of June 2026 — verify with Anthropic for current availability. Last updated 2026-07-23.*

Claude Fable 5 — Files