Google
availableShows if the model has enough results for an index.Gemini 3.8 Flash
Gemini 3.8 Flash is a reasoning model from Google. 47 benchmarks count toward its score, in 7 categories.
IndexOverall score. 50 is the middle.72.1 ±2.9
CoverageShare of the index weight with results.95%
SpeedOutput tokens per second.86/s
Input / 1MUS dollars per 1M input tokens.$0.75 batch $0.375
Output / 1MUS dollars per 1M output tokens.$3.75 batch $1.88 US dollars per 1M output tokens in a batch.
ContextMaximum tokens in one request.1.05M
EloLMArena rating and rank.1495 (#6)
50 is the middle of the board. The range shows the doubt in the index. Batch work costs less.
5,076 votes. Elo shows what people prefer. It does not change the score.
CapabilitiesScore per category. 50 is the middle.
50 is the middleResults
47 counted| BenchmarkThe test name. | CategoryThe capability that the test measures. | ResultThe score from the publisher. | IndexThis result on the index scale. | RunThe settings of the run. | DateDate of the result. | Published byThe source of the result. |
|---|---|---|---|---|---|---|
| OTIS Mock AIME 2024-2025 | Math | 98.9% | 66.8 | high effort | — | Epoch AI |
| GPQA diamond | Knowledge | 95.4% | 66.3 | high effort | — | Epoch AI |
| Artificial Analysis GPQA Diamond | Knowledge | 95.3% | 66.2 | — | — | Artificial Analysis |
| GPQA Diamond | Knowledge | 94.4% | 65.4 | — | 1 Sept 2026 | Vals AI |
| LiveBench Mathematics | Math | 91.6% | 70.2 | high effort | 25 Jun 2026 | LiveBench |
| MMLU Pro | Knowledge | 90.2% | 62.7 | — | 1 Sept 2026 | Vals AI |
| LiveCodeBench | Coding | 89.5% | 65.7 | — | 1 Sept 2026 | Vals AI |
| Terminal-Bench 2.1 (provider run) | Agentic | 89.4% | 76.8 | — | — | DeepSeek-AI |
| Terminal-Bench 2.1 (provider run) | Agentic | 89.4% | 76.8 | — | — | DeepSeek-AI |
| LiveBench Reasoning | Reasoning | 89.3% | 80.0 | high effort | 25 Jun 2026 | LiveBench |
| MMMU Pro | Multimodal | 89.1% | 72.3 | — | 1 Sept 2026 | Vals AI |
| BioMysteryBench Human Solvable | Knowledge | 88.8% | — | — | — | Anthropic |
| LiveBench Language | Knowledge | 87.8% | 76.4 | high effort | 25 Jun 2026 | LiveBench |
| LVBench | Multimodal | 87.1% | — | — | — | Qwen Team |
| CharXiv Reasoning without tools | Multimodal | 86.2% | — | — | — | CharXiv authors |
| LABBench2: An Improved Benchmark for AI Systems Performing Biology Research | Knowledge | 86.2% | — | — | — | Jon M. Laurent et al. |
| Artificial Analysis MMMU-Pro | Multimodal | 85.6% | 71.7 | — | — | Artificial Analysis |
| LiveBench Instruction Following | Instruction | 81.4% | 88.4 | high effort | 25 Jun 2026 | LiveBench |
| Artificial Analysis Long Context Reasoning | Reasoning | 81.3% | 64.5 | — | — | Artificial Analysis |
| Terminal-Bench 2.1 | Agentic | 81.3% | 72.0 | — | 21 Sept 2026 | Vals AI |
| SWE-bench | Coding | 80.0% | 61.7 | — | 1 Sept 2026 | Vals AI |
| Vibe Code Bench v1.1 | Coding | 78.7% | 74.9 | OpenHands | 21 Sept 2026 | Vals AI |
| Artificial Analysis Coding Index | Coding | 76.3% | 72.7 | — | — | Artificial Analysis |
| DeepSWE | Agentic | 73.8% | 77.4 | — | — | Datacurve AI |
| LiveBench Coding | Coding | 72.5% | 58.0 | high effort | 25 Jun 2026 | LiveBench |
| SimpleQA Verified | Knowledge | 69.7% | 86.1 | high effort | — | Epoch AI |
| cursorBench32 | Coding | 69.2% | 77.1 | — | — | Benchmark authors |
| FrontierMath-Tiers-1-3-v2-Private | Math | 68.4% | 70.0 | high effort | — | Epoch AI |
| Chess Puzzles | Reasoning | 61.0% | 95.0 | high effort | — | Epoch AI |
| Artificial Analysis AutomationBench | Agentic | 59.9% | 71.9 | — | — | Artificial Analysis |
| OSWorld 2.0 | Agentic | 59.0% | 84.0 | — | — | Mengqi Yuan et al. |
| SkillsBench | Coding | 58.0% | 72.1 | OpenHands | 11 Sept 2026 | Vals AI |
| IOI | Coding | 56.9% | 70.5 | — | 21 Sept 2026 | Vals AI |
| Artificial Analysis SciCode | Coding | 56.6% | 71.4 | — | — | Artificial Analysis |
| BioMysteryBench Human Difficult | Knowledge | 56.5% | — | — | — | Anthropic |
| HLE-Verified | Knowledge | 54.9% | — | — | — | Weiqi Zhai et al. |
| Artificial Analysis Omniscience Accuracy | Knowledge | 54.6% | 81.4 | — | — | Artificial Analysis |
| LiveBench Agentic Coding | Agentic | 54.2% | 67.8 | high effort | 25 Jun 2026 | LiveBench |
| LiveBench Data Analysis | Reasoning | 54.0% | 30.9 | high effort | 25 Jun 2026 | LiveBench |
| ProofBench v1.1 | Math | 48.0% | 68.8 | — | 21 Sept 2026 | Vals AI |
| Artificial Analysis Humanity's Last Exam | Knowledge | 47.8% | 76.4 | — | — | Artificial Analysis |
| Mystery Game Puzzles | Reasoning | 47.0% | 83.9 | high effort | — | Epoch AI |
| GDPval-AA normalized | Agentic | 45.6% | 70.6 | — | — | Artificial Analysis |
| Artificial Analysis Tau3-Banking | Agentic | 44.9% | 73.6 | — | — | Artificial Analysis |
| Artificial Analysis Agentic Index | Agentic | 41.1% | 70.7 | — | — | Artificial Analysis |
| Artificial Analysis Intelligence Index | Knowledge | 40.9% | 73.6 | — | — | Artificial Analysis |
| cursorBench40 | Coding | 39.6% | — | — | — | Benchmark authors |
| Code Migration | Coding | 36.5% | 68.3 | — | 21 Sept 2026 | Vals AI |
| Furniture Assembly | Reasoning | 31.7% | 62.0 | high effort | — | Epoch AI |
| ApprenticeBench: end-to-end computer use, continual learning, and long-horizon agency on a real accounts-payable job | Agentic | 24.0% | 73.5 | — | — | NeoCognition |
| FrontierMath-Tier-4-v2-Private | Math | 22.0% | 58.5 | high effort | — | Epoch AI |
| Medical Long Context Reasoning (MLCR-AA) | Reasoning | 21.7% | 56.6 | — | — | Wisedocs and Artificial Analysis |
| Artificial Analysis GDP.pdf | Agentic | 21.0% | 72.2 | — | — | Artificial Analysis |
| FrontierSWE v2 | Coding | 19.6% | 65.3 | — | — | Proximal |
| Terminal-Bench 4.0.0 | Agentic | 19.1% | 72.9 | high effort · mini-SWE-agent | 21 Sept 2026 | Terminal-Bench |
| Vibe Code Bench 1-100 | Coding | 18.8% | 72.4 | OpenHands | 16 Sept 2026 | Vals AI |
| Critical Physics Tasks | Reasoning | 18.3% | 76.7 | — | — | Artificial Analysis |
| Terminal-Bench 4.0 | Agentic | 13.1% | 68.7 | — | 21 Sept 2026 | Vals AI |
| Agent Arena task outcome | Agentic | 9.3 | 77.9 | high effort | 15 Sept 2026 | LMArena |
| Agent Arena command recovery | Agentic | 1.9 | 69.5 | high effort | 15 Sept 2026 | LMArena |
| ProgramBench | Coding | 1.0% | — | — | 21 Sept 2026 | Vals AI |
| Agent Arena steerability | Agentic | -1.2 | 66.0 | high effort | 15 Sept 2026 | LMArena |
47 benchmarks count, from 54 of 62 results. A grey row does not count. Too few models took that benchmark.