Matched GPT-5 class at a fraction of the cost, open-weight MoE
| Benchmark | Score | Rank |
|---|---|---|
FrontierMath Unpublished research-level mathematics problems | 2% | #4 / 4 |
MMLU Tests knowledge across 57 subjects from STEM to humanities | 92.6% | #8 / 53 |
HumanEval Coding ability - generating correct Python functions | 95.4% | #8 / 49 |
ARC-C Grade-school science questions requiring reasoning | 98.2% | #8 / 40 |
HellaSwag Common sense reasoning about everyday situations | 96.8% | #8 / 36 |
MATH Competition-level mathematics problems | 96.8% | #11 / 49 |
Arena Elo Human preference ranking via blind comparisons | 1418 | #17 / 41 |
SWE-bench Real-world GitHub issue resolution | 72.8% | #19 / 38 |
GPQA PhD-level science questions even experts struggle with | 86.2% | #20 / 54 |
TerminalArtificial Analysis Agentic terminal coding tasks requiring multi-step execution | 34.8% | #20 / 37 |