Anthropic

The most recent Anthropic model tracked on AI Release Tracker is Claude Fable 5.1, released on Sep 1 2026.

AI Release Tracker tracks 27 Anthropic model releases, from Claude 1 on Mar 14 2023 through Claude Fable 5.1 on Sep 1 2026.

Anthropic's tracked releases, newest first: Claude Fable 5.1 (Sep 1 2026), Claude Mythos 5.1 (Sep 1 2026), Claude Opus 5 (Jul 24 2026), Claude Sonnet 5 (Jun 30 2026), Claude Fable 5 (Jun 9 2026), Claude Mythos 5 (Jun 9 2026), Claude Opus 4.8 (May 28 2026), Claude Opus 4.7 (Apr 16 2026), Claude Sonnet 4.6 (Feb 17 2026), Claude Opus 4.6 (Feb 5 2026), Claude Opus 4.5 (Nov 24 2025), Claude Haiku 4.5 (Oct 15 2025), Claude Sonnet 4.5 (Sep 29 2025), Claude Opus 4.1 (Aug 5 2025), Claude Sonnet 4 (May 22 2025), Claude Opus 4 (May 22 2025), Claude 3.7 Sonnet (Feb 24 2025), Claude 3.5 Haiku (Oct 22 2024), Claude 3.5 Sonnet (upgraded) (Oct 22 2024), Claude 3.5 Sonnet (Jun 20 2024), Claude 3 Haiku (Mar 4 2024), Claude 3 Sonnet (Mar 4 2024), Claude 3 Opus (Mar 4 2024), Claude 2.1 (Nov 21 2023), Claude Instant 1.2 (Aug 9 2023), Claude 2 (Jul 11 2023), Claude 1 (Mar 14 2023).

Among all models tracked here, Anthropic holds the best score on BullshitBench v2 (Claude Opus 4.8, 95%), Gray Swan IPI (k = 1) (Claude Opus 5, 0.2%), Gray Swan IPI (k = 10) (Claude Opus 5, 1.6%), Gray Swan IPI (k = 15) (Claude Opus 5, 2%), ProgramBench (Claude Opus 5, 4.5%), SWE-Bench Pro (Claude Fable 5.1, 81.2%), SWE-Bench Verified (Claude Opus 5, 96%), SWE-Bench Multilingual (Claude Opus 5, 89.5%), SWE-Bench Multimodal (Claude Opus 5, 59.4%), FrontierCode v1.1 (Main) (main split) (Claude Opus 5, 53.4%), Next.js Evals (Claude Fable 5.1, 97%), Supabase Evals (with skills) (Claude Sonnet 5, 91.3%), Supabase Evals (no skills) (Claude Opus 5, 89.9%), Frontier-Bench v0.1 (Claude Opus 5, 43.3%), Terminal-Bench 4.0 (Claude Mythos 5.1, 60.9%), Terminal-Bench 2.0 (Claude Mythos 5, 88%), BU Bench (Claude Opus 4.8, 74%), Humanity's Last Exam (no tools) (Claude Fable 5.1, 60.9%), Humanity's Last Exam (with tools) (Claude Fable 5.1, 65%), BioMysteryBench (human solved) (Claude Opus 5, 90.1%), OSWorld 2.0 (Claude Opus 5, 70.6%), Harvey's Legal Agent Benchmark (Held-out) (Claude Opus 5, 11.7%), HealthBench Professional (Claude Fable 5, 63.3%), GDPval-AA (Claude Fable 5, 1932), GDPval-AA v2 (Claude Opus 5, 1861) and AA-Briefcase (Claude Fable 5.1, 1694).

No. All 27 tracked Anthropic releases are proprietary — available through Anthropic's own API or products rather than as downloadable weights.

Across its 27 tracked releases, Anthropic has shipped a new model roughly every 49 days on average — about 7.4 releases per year.