B.O.B. Leaderboard
Google DeepMindreleased 2 Sep 2026
Gemini 3.8 Flash
#19 of 9534 of 46 benchmarksConsistent
64.1
B.O.B.
Category profile
- Real use (hard to game)2×63
- Reasoning & knowledge81
- Math52
- Coding59
- Agents & real work59
- Independent indexes71
Reality check
Lab tests
63
Real use (7 tests)
63
Lab tests and real-use tests agree. Its lab-test and real-use scores are within 2 points of each other.
What counts as real usePrice per 1M tokens (OpenRouter)
$0.75 in · $3.75 out
Cost per task on the AA index
$1.24
Output speed (Artificial Analysis)
145 tok/s
§Receipts
Every benchmark result
“Beats” = the share of tracked models this one beat on that test. The raw score is its best reported setting.
| Benchmark | Raw | Beats | Source |
|---|---|---|---|
| Real use (hard to game) · Real use 63 | |||
| LMArena Textas gemini-3.8-flash-high | 1496 | 91%of 80 | LMArena (blind human votes) |
| LMArena Agent Arenaas Gemini 3.8 Flash (High) | 0.040 | 76%of 50 | LMArena (real agent sessions) |
| LMArena WebDevas gemini-3.8-flash-high | 1585 | 67%of 82 | LMArena (blind human votes) |
| Vals Indexas google/gemini-3.8-flash | 54.9% | 63%of 42 | Vals AI (private test sets) |
| BridgeBench (overall)as Gemini 3.8 Flash | 584 | 61%of 29 | BridgeMind |
| EQ-Bench Creative Writingas gemini-3.8-flash | 1748 | 47%of 46 | Sam Paech (open source) |
| LiveBench 2.0 (real-work suites)as gemini-3.8-flash | 65.6% | 40%of 39 | LiveBench / Abacus.AI |
| Reasoning & knowledge · Reasoning 81 | |||
| GPQA Diamondas gemini-3.8-flash_high | 95.4% | 95%of 70 | Epoch AI Benchmarking Hub |
| SimpleQA Verifiedas gemini-3.8-flash_high | 69.7% | 87%of 57 | Epoch AI Benchmarking Hub |
| Chess Puzzlesas gemini-3.8-flash_high | 58.0% | 86%of 73 | Epoch AI Benchmarking Hub |
| ARC-AGI-2as gemini-3.8-flash_high | 89.2% | 83%of 44 | Epoch AI Benchmarking Hub |
| Blueprint-Bench 2as gemini-3.8-flash_unknown | 38.6% | 82%of 31 | Epoch AI Benchmarking Hub |
| Mystery Game Puzzlesas gemini-3.8-flash_high | 47.0% | 80%of 57 | Epoch AI Benchmarking Hub |
| CritPt (physics research)as gemini-3.8-flash_high | 18.3% | 67%of 89 | Epoch AI Benchmarking Hub |
| Humanity's Last Examas gemini-3.8-flash_unknown | 44.5% | 67%of 13 | Epoch AI Benchmarking Hub |
| Math · Math 52 | |||
| OTIS Mock AIMEas gemini-3.8-flash_high | 98.9% | 79%of 73 | Epoch AI Benchmarking Hub |
| FrontierMath Tiers 1-3as gemini-3.8-flash_high | 68.4% | 58%of 60 | Epoch AI Benchmarking Hub |
| ProofBenchas gemini-3.8-flash_unknown | 48.0% | 51%of 68 | Epoch AI Benchmarking Hub |
| FrontierMath Tier 4as gemini-3.8-flash_high | 22.0% | 23%of 52 | Epoch AI Benchmarking Hub |
| Coding · Coding 59 | |||
| DeepSWEas gemini-3.8-flash_high | 73.8% | 96%of 26 | Epoch AI Benchmarking Hub |
| WeirdMLas gemini-3.8-flash_high | 84.8% | 84%of 52 | Epoch AI Benchmarking Hub |
| SciCodeas gemini-3.8-flash_high | 56.6% | 78%of 86 | Epoch AI Benchmarking Hub |
| ALE-Benchas gemini-3.8-flash_high | 1269.7% | 63%of 71 | Epoch AI Benchmarking Hub |
| FrontierCodeas gemini-3.8-flash_medium | 41.2% | 46%of 45 | Epoch AI Benchmarking Hub |
| CursorBenchas gemini-3.8-flash_high | 39.6% | 25%of 17 | Epoch AI Benchmarking Hub |
| FrontierSWEas gemini-3.8-flash_high | 19.6% | 21%of 20 | Epoch AI Benchmarking Hub |
| Agents & real work · Agents 59 | |||
| APEX-Agentsas gemini-3.8-flash_unknown | 64.3% | 83%of 48 | Epoch AI Benchmarking Hub |
| DTBenchas gemini-3.8-flash_unknown | 95.7% | 76%of 67 | Epoch AI Benchmarking Hub |
| LMCAas gemini-3.8-flash_medium | 52.9% | 76%of 67 | Epoch AI Benchmarking Hub |
| GDP.pdfas gemini-3.8-flash_medium | 23.4% | 56%of 37 | Epoch AI Benchmarking Hub |
| Vending-Bench 2as gemini-3.8-flash_unknown | 5093.8% | 37%of 47 | Epoch AI Benchmarking Hub |
| Remote Labor Indexas gemini-3.8-flash_unknown | 5.8% | 25%of 9 | Epoch AI Benchmarking Hub |
| Independent indexes · Indexes 71 | |||
| Epoch Capabilities Indexas Gemini 3.8 Flash | 157.2 | 78%of 81 | Epoch AI |
| Artificial Analysis Intelligence Indexas Gemini 3.8 Flash (high) | 41 | 65%of 54 | Artificial Analysis |