Claude Fable 5.1 holds #1: 1M-token input, ties GPT 6 Astra on new benchmarks
DeepLearningAI · x · 2026-09-15
DeepLearning.AI's The Batch breaks down Anthropic's Claude Fable 5.1: Fable and Claude Mythos 5.1 are the same model differing only in safeguards — Fable is open to everyone, Mythos restricted to US cybersecurity/life-sciences orgs via Project Glasswing, with fallback to Opus 4.8/5 when bio/cyber classifiers trigger.
- Specs: text+image input up to 1M tokens, output up to 128k tokens at 69 tokens/sec, June 2026 knowledge cutoff
- Features: always-on reasoning with five levels (switchable mid-conversation without invalidating cache, beta), tool use with readable progress, prompt caching, per-turn system messages, statistical watermarking
- Scores: ties for first on Artificial Analysis Intelligence Index v4.3 (53), leads Vals AI (68.83%), scores 57 on v4.2 edging out GPT 6 Astra, 52.6% on Terminal Bench Science 0.1
- Cost: cheaper cache reads for repeated agentic workflows; available via Claude.ai, API, AWS, Google Cloud, Azure
More from Models
- Yale physicists re-grade AI benchmarks: most are broken, frontier models near saturation — inductionheads · 2026-09-15
- Grok-5 Shipping Soon, Says xAI Isn't Slowing Down as It Chases the Frontier — bindureddy · 2026-09-15
- Podcast: Microsoft researcher hits ~45% on ARC-AGI with tiny recursive networks — ziv_ravid · 2026-09-15
- Users report being routed to GPT-6 Sol, early impressions consistently positive — kimmonismus · 2026-09-15
- Tests show OpenAI hasn't changed Codex quotas: 800M+ Astra tokens per cycle — daniel_mac8 · 2026-09-15
- AlphaSense tests Fable and Astra as financial researchers: top score but only on par with Opus-5 — CShorten30 · 2026-09-15