Claude Mythos 5
Last updated: June 10, 2026
Specs
| Vendor | Anthropic |
|---|
| Released | June 9, 2026 |
|---|
| Context window | 1M tokens |
|---|
| Input | $10 / M tokens |
|---|
| Output | $50 / M tokens |
|---|
Benchmarks
| mmlu | 95.1 |
|---|
| gpqa | 92 |
|---|
| humanEval | 98 |
|---|
| arenaElo | 1493 |
|---|
| sweBench | 78 |
|---|
| sweBenchPro | 73 |
|---|
| liveCodeBench | 85 |
|---|
| aiderPolyglot | 87 |
|---|
| terminalBench | 53 |
|---|
| mmluPro | 89.5 |
|---|
| hle | 31 |
|---|
| aime | 94.5 |
|---|
| math500 | 98.8 |
|---|
| mmmu | 83.5 |
|---|
Strengths
- Same frontier capabilities, pricing and API surface as Claude Fable 5, Anthropic's most capable model
- 1M-token context window (the default) with 128K max output and always-on protected thinking
- Delivered through Project Glasswing, succeeding the invitation-only Claude Mythos Preview
Weaknesses
- Access limited to Project Glasswing participants: there is no general self-serve API
- Same premium pricing ($10 / $50 per M tokens) and ~30% heavier new tokenizer as Fable 5
- Requires 30-day data retention (not available under zero-data-retention); safety classifiers may refuse some bio/cyber-adjacent requests
Best for
- Project Glasswing partners that need the most capable Claude available
- Frontier reasoning, research and long-horizon autonomous agents
- Workloads that justify gated frontier access over the generally-available Fable 5
Compare with other AI models