---
title: "Claude Mythos Benchmarks - RankedAGI"
description: "Detailed benchmark and metadata record for Claude Mythos."
source: "https://rankedagi.com/models/claude-mythos-preview"
---

# Claude Mythos

| Field | Value |
| --- | --- |
| Organization | Anthropic |
| License | Proprietary |
| Type | Thinking |
| Version | Preview |
| Released | 2026-04-07 |
| Input cost per million tokens | $25 |
| Output cost per million tokens | $125 |
| Last updated | 2026-04-20T06:05:16.239Z |

## Benchmarks

| Benchmark | Category | Value | Description | Source |
| --- | --- | --- | --- | --- |
| RankedAGI Coding | coding | 95.2% | RankedAGI Coding Score |  |
| SWEBench Pro | coding | 77.8% | Diverse Agentic Coding |  |
| SWEBench Multilingual | coding | 87.3% |  |  |
| SWEBench Verified | coding | 93.9% | Agentic Coding | https://www.swebench.com/index.html |
| Terminal-Bench 2.0 | coding | 82.0% | Agentic Terminal Coding | https://www.tbench.ai/leaderboard/terminal-bench/2.0 |
| RankedAGI Agentic | Agents | 93.5% | RankedAGI Agentic Score |  |
| BrowseComp | Agents | 86.9% | A benchmark for browsing agents | https://openai.com/index/browsecomp/ |
| OSWorld-Verified | Agents | 79.6% |  |  |
| CyberGym | safety | 83.1% |  |  |
| RankedAGI Reasoning | reasoning | 89.5% | RankedAGI Reasoning Score |  |
| Humanity's Last Exam | reasoning | 56.8% | Multidisciplinary Reasoning (no tools) |  |
| Humanity's Last Exam w/ Tools | reasoning | 64.7% | Multidisciplinary Reasoning (with tools) |  |
| GPQA Diamond | reasoning | 94.6% | Generalized Prefix Question Answering Score (Reasoning) PhD Level Reasoning |  |
| RankedAGI Math | math | 71.5% | RankedAGI Math Score |  |
| RankedAGI Overall | general | 78.8% | Overall RankedAGI score |  |
| CharXiv Reasoning | Multimodal | 86.1% | Figure Understanding |  |
| CharXiv Reasoning w/ Tools | Multimodal | 93.2% | Figure Understanding |  |

## Model Links

| Title | URL |
| --- | --- |
| Release | https://www.anthropic.com/glasswing |