Atlas

Models

← All models

Claude Mythos Preview

Anthropic · 2026-04-07

Claude Mythos Preview is a general-purpose frontier model with advanced agentic coding and reasoning skills and a major increase in cybersecurity capability over prior Claude models. Anthropic made it available only to invited Project Glasswing partners for defensive cybersecurity work; on the API, adaptive thinking is the default and thinking depth can be varied with the effort parameter.

Benchmark scores

BenchmarkScore
BrowseComp87.9
CAISI PortBench80.1
Capability164.2
CharXiv Reasoning86.1
CharXiv Reasoning (with tools)93.2
CTF-Archive Diamond65.6
Cybench Anthropic 35-task repeated-trial subset100.0
CyberGym83.1
DeepSearchQA (F1 score)94.4
ExploitBench (CAISI Inspect + AutoNudge)57.2
ExploitBench Leaderboard (Capability coverage)78.0
ExploitBench Leaderboard (Mean capability score, out of 16)10.0
FrontierScience Olympiad83.0
GPQA Diamond92.9
GraphWalks BFS - 1M74.3
GraphWalks BFS - 256K85.7
GraphWalks BFS - card aggregate over 256K and 1M80.0
Humanity's Last Exam56.8
Humanity's Last Exam (with tools)64.7
MMMLU92.7
OSWorld-Verified79.6
OSWorld-Verified (Anthropic 361-task cohort)85.4
OTIS-AIME 2025 (CAISI 30-task set)99.5
PUMaC 2024 (CAISI 55-task set)97.7
RiemannBench (Anthropic no-tools protocol)43.0
SMT 2025 (CAISI 58-task set)95.0
SWE-bench Multilingual87.3
SWE-bench Multimodal59.0
SWE-Bench Pro (Public Dataset)77.8
SWE-bench Verified93.9
Loading Atlas data…