Anthropic Launches Claude Fable 5.1 and Mythos 5.1: Benchmarks, Pricing, and Access

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026, calling them its most advanced models for coding and knowledge work. Fable 5.1 succeeds Fable 5 as the most capable widely released Claude model at the same $10/$50 per million token price; Mythos 5.1 is the same model with relaxed safeguards, offered by invitation to vetted US organizations.

What’s New

• Beats Opus 5 on every published benchmark — including 55.8% vs 52.3% on Terminal-Bench 4.0, 73.4% vs 70.0% on CursorBench 3.2, and 52.6% vs 29.0% on Terminal-Bench-Science, a new agentic scientific-research evaluation. Full table in the Opus 5 vs Fable 5.1 comparison.
• Cache reads cut 75% to $0.25 per million tokens. Anthropic estimates typical workloads run about 25% cheaper than on Fable 5, and highly agentic work up to roughly 45% cheaper.
• Newest knowledge cutoff of any Claude model: June 2026.
• Fewer false refusals: about 60% fewer cybersecurity false positives and 85% fewer biology safeguard triggers on benign requests, and the model now permits discovering software vulnerabilities.
• New API controls: per-message effort, turn-scoped system messages, and readable progress updates between tool calls (all beta) — plus three breaking changes for existing Fable 5 integrations, chiefly that forced tool use now returns an error.
• Science results: Anthropic reports Mythos 5.1 designed protein binders with roughly 10× higher binding affinity than competition winners and a near-50% hit rate across 12 targets, delivered up to 2.5× speedups in computational biology, and mapped Venus at 2–3 km resolution from reprocessed radar data.

Availability

Fable 5.1 is available to all customers from day one on the Claude API (claude-fable-5-1), Amazon Bedrock, Google Cloud, Microsoft Foundry, Claude Code, and Claude Enterprise. Fable 5 moves to legacy status but stays available until at least June 9, 2027. Mythos 5.1 access goes through Project Glasswing’s Cyber Verification and Life Sciences Verification programs. Anthropic’s guidance for choosing: start with Claude Opus 5 for most workloads, and use Fable 5.1 for demanding reasoning and long-horizon agentic work.

Alongside the models, Anthropic previewed Enterprise Frontier Safeguards, letting enterprises run frontier-model safeguards on their own cloud infrastructure, rolling out in phases this fall. Fable 5.1’s text also carries Anthropic’s statistical watermark on every platform.

Frequently Asked Questions