Natively multimodal with voice, 2x faster and 50% cheaper than GPT-4 Turbo
SI v3 uses weighted percentages from one Artificial Analysis configuration. Ranking requires all five benchmarks. Partial scores average only available weights. Scores are not comparable to SI v2 or a probability of reaching singularity.
Compared with 9 models released within six months of it, across 3 benchmarks where scores spread out.
| Benchmark | Score | Rank |
|---|---|---|
HellaSwag Common sense reasoning about everyday situations | 95.3% | #13 / 36 |
ARC-C Grade-school science questions requiring reasoning | 96.7% | #22 / 40 |
MMLU Tests knowledge across 57 subjects from STEM to humanities | 88.7% | #27 / 54 |
HumanEval Coding ability - generating correct Python functions | 90.2% | #31 / 50 |
MATH Competition-level mathematics problems | 76.6% | #36 / 50 |
SWE-bench Real-world GitHub issue resolution | 33.2% | #39 / 39 |
Arena Elo Human preference ranking via blind comparisons | 1285 | #42 / 51 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 56.6% | #50 / 51 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 62.7% | #61 / 61 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 26.4% | #62 / 62 |
GPQA PhD-level science questions even experts struggle with | 53.6% | #88 / 96 |
HLEArtificial Analysis Challenging multidisciplinary questions evaluated by Artificial Analysis | 4.2% | #95 / 100 |
| Benchmark / source | Score |
|---|---|
HLE Artificial AnalysisGPT-4o mini | 4.2% |
MMMU-Pro Artificial AnalysisGPT-4o mini | 41.5% |
Terminal-Bench 2.1 Artificial AnalysisGPT-4o mini | 5.6% |
IFBench Artificial AnalysisGPT-4o mini | 31% |