| Claude Mythos | Grok 4 Heavy | |
|---|---|---|
| arenaElo | 1478 | 1391 |
| humanEval | 98.7 | 88.4 |
| sweBench | 93.9 | 55 |
| mmlu | 94.2 | 89.3 |
| gpqa | 94.6 | 74.5 |
| aime | 97.6 | 92 |
| Price in/out ($/M) | $25 / $125 | $15 / $60 |
Grok 4 Heavy uses multi-agent debate to hit 74.5% GPQA at $15 / $60 per M tokens. Mythos posts 94.6% GPQA in a single forward pass at $25 / $125. On Humanity's Last Exam with tools Mythos leads by ~6 points. For most teams Grok 4 Heavy is the pragmatic ceiling. Mythos is the answer when you need maximum capability and have access.