B.O.B. Leaderboard
Xiaomireleased 21 Sep 2026
MiMo-V2.6-Pro
#14 of 9512 of 46 benchmarksConsistent
75.5
B.O.B.
Category profile
- Real use (hard to game)2×69
- Reasoning & knowledge83
- Math76
- Coding74
- Agents & real work75
- Independent indexes82
Reality check
Lab tests
77
Real use (6 tests)
69
Lab tests and real-use tests agree. It does 8 points better on the usual lab tests than on tests that are hard to train for.
What counts as real usePrice per 1M tokens (OpenRouter)
$0.43 in · $0.87 out
Cost per task on the AA index
$0.13
Output speed (Artificial Analysis)
46 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 69 | |||
| LMArena WebDevas mimo-v2.6-pro | 1625 | 80%of 82 | LMArena (blind human votes) |
| LMArena Agent Arenaas MiMo V2.6 Pro | 0.038 | 74%of 50 | LMArena (real agent sessions) |
| LMArena Textas mimo-v2.6-pro | 1479 | 72%of 80 | LMArena (blind human votes) |
| Vals Indexas xiaomi/mimo-v2.6-pro | 55.2% | 71%of 42 | Vals AI (private test sets) |
| LiveBench 2.0 (real-work suites)as mimo-v2.6-pro | 74.4% | 66%of 39 | LiveBench / Abacus.AI |
| BridgeBench (overall)as MiMo-V2.6-Pro | 568 | 54%of 29 | BridgeMind |
| Reasoning & knowledge · Reasoning 83 | |||
| CritPt (physics research)as mimo-v2.6-pro | 26.6% | 83%of 89 | Epoch AI Benchmarking Hub |
| Math · Math 76 | |||
| ProofBenchas mimo-v2.6-pro | 70.0% | 76%of 68 | Epoch AI Benchmarking Hub |
| Coding · Coding 74 | |||
| SciCodeas mimo-v2.6-pro | 60.9% | 94%of 86 | Epoch AI Benchmarking Hub |
| ALE-Benchas mimo-v2.6-pro | 1158.1% | 54%of 71 | Epoch AI Benchmarking Hub |
| Agents & real work · Agents 75 | |||
| APEX-Agentsas mimo-v2.6-pro | 59.5% | 75%of 48 | Epoch AI Benchmarking Hub |
| Independent indexes · Indexes 82 | |||
| Artificial Analysis Intelligence Indexas MiMo-V2.6-Pro | 46 | 82%of 54 | Artificial Analysis |