Flash-tier model with near-frontier reasoning: GPQA 92.2, ARC-AGI-1 96%. AA Intelligence Index 55.3 — top of the cost-efficient tier.
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 models released within six months of it.
| Benchmark | Score | Rank |
|---|---|---|
Arena EloArtificial Analysis Human preference ranking via blind comparisons | 1655 | #3 / 51 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 88.3% | #9 / 51 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 89.5% | #10 / 61 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 87.6% | #12 / 62 |
ARC-AGIARC Prize Novel reasoning tasks requiring fluid intelligence | 72.1% | #16 / 40 |
GPQAArtificial Analysis PhD-level science questions even experts struggle with | 92.2% | #26 / 96 |
HLEArtificial Analysis Challenging multidisciplinary questions evaluated by Artificial Analysis | 42.7% | #30 / 100 |
TerminalArtificial Analysis Agentic terminal coding tasks requiring multi-step execution | 40.9% | #31 / 58 |
| Benchmark / source | Score |
|---|---|
HLE Artificial AnalysisGemini 3.5 Flash (High) | 42.7% |
MMMU-Pro Artificial AnalysisGemini 3.5 Flash (High) | 84.3% |
Terminal-Bench 2.1 Artificial AnalysisGemini 3.5 Flash (High) | 78.7% |
Terminal-Bench 4.0 Artificial AnalysisGemini 3.5 Flash (High) | 6.6% |
SciCode Artificial AnalysisGemini 3.5 Flash (High) | 53.9% |
AA-LCR Artificial AnalysisGemini 3.5 Flash (High) | 73.3% |
CritPt Artificial AnalysisGemini 3.5 Flash (High) | 13.1% |
ITBench SRE Artificial AnalysisGemini 3.5 Flash (High) | 40.3% |
IFBench Artificial AnalysisGemini 3.5 Flash (High) | 76.3% |