Incremental upgrade with improved reliability and instruction following
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 38 models released within six months of it, across 8 benchmarks where scores spread out.
| Benchmark | Score | Rank |
|---|---|---|
HellaSwag Common sense reasoning about everyday situations | 97.8% | #3 / 36 |
ARC-C Grade-school science questions requiring reasoning | 98.4% | #5 / 40 |
HumanEval Coding ability - generating correct Python functions | 95.6% | #7 / 50 |
MATH Competition-level mathematics problems | 97.8% | #7 / 50 |
MMLU Tests knowledge across 57 subjects from STEM to humanities | 92.4% | #9 / 54 |
SWE-bench Real-world GitHub issue resolution | 76.3% | #13 / 39 |
TerminalArtificial Analysis Agentic terminal coding tasks requiring multi-step execution | 45.5% | #21 / 58 |
Arena Elo Human preference ranking via blind comparisons | 1425 | #25 / 51 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 85.6% | #25 / 62 |
ARC-AGIARC Prize Novel reasoning tasks requiring fluid intelligence | 18.3% | #33 / 40 |
GPQA PhD-level science questions even experts struggle with | 88.1% | #52 / 96 |
HLEArtificial Analysis Challenging multidisciplinary questions evaluated by Artificial Analysis | 28.5% | #69 / 100 |
| Benchmark / source | Score |
|---|---|
HLE Artificial AnalysisGPT-5.1 (High) | 28.5% |
MMMU-Pro Artificial AnalysisGPT-5.1 (High) | 75.5% |
Terminal-Bench 2.1 Artificial AnalysisGPT-5.1 (High) | 52.4% |
AA-LCR Artificial AnalysisGPT-5.1 (High) | 80% |
CritPt Artificial AnalysisGPT-5.1 (High) | 4.9% |
IFBench Artificial AnalysisGPT-5.1 (High) | 72.9% |