New12,480+ smartphones indexed — free sandbox tier on every plan.Edge-cached JSON responses, typically under 100ms.SDKs for C#, TypeScript and Python, plus an MCP server for AI agents.New12,480+ smartphones indexed — free sandbox tier on every plan.Edge-cached JSON responses, typically under 100ms.SDKs for C#, TypeScript and Python, plus an MCP server for AI agents.
Get Free API Key
Artificial Intelligence & Frontier Models•4 min read•2026-11-10

Anthropic Launches Claude Fable 5.2 and Mythos 5.2 Frontier Models for Advanced Reasoning and Sciences

Anthropic unveils Claude Fable 5.2 and Mythos 5.2, delivering 93.0% on SWE-bench Pro, 49.1% on FrontierCode, and a 20% price cut to $8/$40 per million tokens across a 1M context window.

Tech News Bot
DeviceSpecs Intelligence
Original Source Reference
#Anthropic#Claude Fable 5.2#Claude Mythos 5.2#Frontier AI#AI Agents#SWE-bench Pro#Computer Use#Reasoning Models

Anthropic has officially launched Claude Fable 5.2 and Claude Mythos 5.2, its latest flagship foundation models engineered to tackle the most demanding technical challenges across software engineering, quantitative finance, and scientific discovery. The dual-tier release introduces state-of-the-art results across coding, computer use, and formal verification benchmarks alongside a 20% price reduction across its 1-million-token context window.

Dual Tier Deployment: General Availability Fable 5.2 and Verified Mythos 5.2

The launch ships the same underlying model architecture in two distinct governance tiers. Claude Fable 5.2 is generally available today across Claude apps, Claude Code, the Messages API, and Google Cloud Model Garden, incorporating rigorous universal safeguards for cyber and biosecurity domains. Claude Mythos 5.2 provides identical underlying capabilities with specialized security and life science configurations, accessible exclusively to vetted organizations through Anthropic Cyber Verification and Life Sciences Verification programs.

Frontier Technical Benchmarks: SWE-bench Pro, Computer Use, and Formal Proofs

Fable 5.2 establishes new performance ceilings across challenging evaluation suites, scoring 93.0% on SWE-bench Pro (up from 81.2% on Fable 5.1) and a record 49.1% on FrontierCode 2.0 at extra-high effort. On the Remote Labor Index, which benchmarks end-to-end execution of professional freelance software contracts, Fable 5.2 resolved 41.3% of tasks at human professional parity, nearly double Opus 5.5 baseline. In computer use, the model achieved 89.3% on OSWorld 2.1, while a multi-agent swarm of 1,000 model instances formalized a complete Lean proof of Fermat Last Theorem in under 30 hours during pre-release testing.

Architectural Speedups, 20% Price Cut, and Agentic Multi-Model Workflows

Anthropic has reduced API pricing by 20% compared to Fable 5.1, setting rates at $8.00 per million input tokens and $40.00 per million output tokens, with prompt cache reads priced at $0.20 per million tokens. The release introduces a dedicated fast mode delivering up to 30% latency reductions on the Claude API, native support for computer use tools, and an advisor framework where Fable 5.2 and Mythos 5.2 agents can orchestrate, verify, and cross-evaluate subordinate execution agents across complex long-horizon workflows.