B.O.B. Leaderboard
StepFunreleased 18 Sep 2026
Step 5 Preview
#26 of 959 of 46 benchmarksBenchmaxxed?
61.6
B.O.B.
Category profile
- Real use (hard to game)2×43
- Reasoning & knowledge75
- Math46
- Coding87
- Agents & real work—
- Independent indexes75
Reality check
Lab tests
70
Real use (5 tests)
43
Scores much higher on lab tests than on real-use tests. It does 26 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)
$1.00 in · $2.70 out
Cost per task on the AA index
$1.03
Output speed (Artificial Analysis)
91 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 43 | |||
| LMArena WebDevas step-5-preview-high | 1564 | 62%of 82 | LMArena (blind human votes) |
| BridgeBench (overall)as Step 5 Preview | 572 | 57%of 29 | BridgeMind |
| LMArena Textas Step 5 Preview | 1446 | 33%of 80 | LMArena (blind human votes) |
| LMArena Agent Arenaas Step 5 Preview | -0.016 | 33%of 50 | LMArena (real agent sessions) |
| Vals Indexas stepfun/step-5-preview | 49.4% | 32%of 42 | Vals AI (private test sets) |
| Reasoning & knowledge · Reasoning 75 | |||
| CritPt (physics research)as step-5-preview_unknown | 20.9% | 75%of 89 | Epoch AI Benchmarking Hub |
| Math · Math 46 | |||
| ProofBenchas step-5-preview_unknown | 42.0% | 46%of 68 | Epoch AI Benchmarking Hub |
| Coding · Coding 87 | |||
| SciCodeas step-5-preview_unknown | 58.9% | 87%of 86 | Epoch AI Benchmarking Hub |
| Independent indexes · Indexes 75 | |||
| Artificial Analysis Intelligence Indexas Step 5 Preview | 44 | 75%of 54 | Artificial Analysis |