Anthropic's Fable 5.1 gains +27.9pp on agentic research but +1.2pp on HLE, cuts price up to 45%
koltregaskes · x · 2026-09-03
Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 — the same model with different safeguard levels, the latter restricted to trusted-access programs for cybersecurity and life sciences. Key details:
- Uneven gains: all eight matched evaluations improved, but wildly unevenly — +27.9pp on agentic scientific research vs just +1.2pp on Humanity's Last Exam with tools.
- Price cuts: 25% cheaper than Fable 5 for typical token-billed workloads (mostly via reduced cache-read pricing), and up to 45% savings for highly agentic use.
- Enterprise Frontier Safeguards (EFS): data stored in customer-controlled cloud infra, delivering zero-data-retention privacy while keeping state-of-the-art adversarial-use defenses; rolling out to enterprises this fall.
- Fewer false positives: cybersecurity guardrails now block 60% fewer benign flags.
More from Models
- Leak: Astra won't be the best model of the year; a 'monster' is slated for end of year — ChrisGPT · 2026-09-03
- Startup Mostik bridges AI models via their weights, tops ARC-AGI 3 at 1/20 the cost — nordicinst · 2026-09-03
- Anthropic launches browser tool to detect Claude-made files via C2PA content credentials — btibor91 · 2026-09-03
- ByteDance's looped language models match 12B rivals at 1.4B size, with Bengio as co-author — peterjliu · 2026-09-03
- Anthropic Weakened Safety Filters, Signed an AI Cyberattack Warning Letter, Then Shipped Mythos 5.1 Anyway — AgentBlackVeil · 2026-09-03
- Marin 535B A23B Frontier-Scale Training Run Is Fully Livestreamed — Sentdex · 2026-09-03