Mythos-class frontier model with cyber/bio safeguards; cache reads cut 75% to $0.25.
| Benchmark | Score | Δ vs prior | Rank (current) | Configuration | Source |
|---|---|---|---|---|---|
Terminal-Bench 2.1 | 85% | #3 of 20 | leaderboard | benchlm.ai ↗ | |
Humanity's Last Exam | 60.9% | #2 of 14 | no tools · lab-reported | vellum.ai ↗ | |
ARC-AGI-2 | 90% | #4 of 9 | leaderboard | benchlm.ai ↗ |
| Version | Released | $ in / out | Terminal-Bench | SWE-bench | GPQA | HLE | ARC-AGI-2 | AIME |
|---|---|---|---|---|---|---|---|---|
| Claude 3.5 Sonnet (Oct 2024) | $3 / $15 | — | 49% | — | — | — | — | |
| Claude 3.7 Sonnet | $3 / $15 | — | 62.3% | — | — | — | — | |
| Claude Opus 4 | $15 / $75 | — | 72.5% | — | — | — | — | |
| Claude Sonnet 4 | $3 / $15 | — | 72.7% | — | — | — | — | |
| Claude Opus 4.1 | $15 / $75 | — | 74.5% | — | — | — | — | |
| Claude Sonnet 4.5 | $3 / $15 | — | 77.2% | — | — | 13.6% | — | |
| Claude Haiku 4.5 | $1 / $5 | 43.8% | 73.3% | — | 10.2% | — | — | |
| Claude Opus 4.5 | $5 / $25 | — | 80.9% | — | 30.8% | 37.6% | 95.1% | |
| Claude Opus 4.6 | $5 / $25 | — | 80.8% | 89.2% | 53% | 68.8% | — | |
| Claude Sonnet 4.6 | $3 / $15 | 57.3% | 79.6% | — | 49% | 58.3% | — | |
| Claude Opus 4.7 | $5 / $25 | 68.5% | 87.6% | 94.2% | 46.9% | 75.8% | — | |
| Claude Opus 4.8 | $5 / $25 | 71.9% | 88.6% | 93.6% | 57.9% | 72.1% | — | |
| Claude Fable 5 | $10 / $50 | 80.5% | 95% | — | 57.8% | 89.2% | — | |
| Claude Mythos 5 | $10 / $50 | — | 95.5% | — | 64.5% | — | — | |
| Claude Sonnet 5 | $2 / $10 | 74.5% | 85.2% | — | 57.4% | — | — | |
| Claude Opus 5 | $5 / $25 | 84.6% | 96% | — | 56.6% | 90.4% | — | |
| Claude Fable 5.1 | $10 / $50 | 85% | — | — | 60.9% | 90% | — | |
| Claude Mythos 5.1 | $10 / $50 | — | — | — | — | — | — | |
| Claude Opus 5.5 | $4 / $20 | 87.6% | — | — | 67.7% | 91.7% | — | |
| Claude Sonnet 5.5 | $2 / $10 | 83.1% | — | — | 56.9% | — | — | |
| Claude Haiku 5.5 | $0.1 / $0.5 | — | — | — | 45.9% | — | — |