B.O.B. Leaderboard
OpenAIreleased 9 Jul 2026
GPT-5.6 Luna
#43 of 9533 of 46 benchmarksBenchmaxxed?
49.0
B.O.B.
Category profile
- Real use (hard to game)2×31
- Reasoning & knowledge47
- Math72
- Coding52
- Agents & real work40
- Independent indexes70
Reality check
Lab tests
53
Real use (10 tests)
31
Scores much higher on lab tests than on real-use tests. It does 22 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)
$0.20 in · $1.20 out
Cost per task on the AA index
—
Output speed (Artificial Analysis)
—
§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 31 | |||
| EQ-Bench Creative Writingas gpt-5.6-luna | 1829 | 58%of 46 | Sam Paech (open source) |
| LMArena WebDevas gpt-5.6-luna-xhigh (codex-harness) | 1518 | 47%of 82 | LMArena (blind human votes) |
| Vals Indexas openai/gpt-5.6-luna | 51.7% | 46%of 42 | Vals AI (private test sets) |
| LMArena Textas gpt-5.6-luna-xhigh | 1453 | 41%of 80 | LMArena (blind human votes) |
| LiveBench 2.0 (real-work suites)as gpt-5.6-luna | 59.8% | 32%of 39 | LiveBench / Abacus.AI |
| LMArena Agent Arenaas GPT 5.6 Luna (xHigh) | -0.018 | 31%of 50 | LMArena (real agent sessions) |
| SWE-rebenchas GPT-5.6 Luna [medium] | 43.6% | 18%of 34 | Nebius |
| BridgeBench (overall)as GPT-5.6 Luna | 476 | 14%of 29 | BridgeMind |
| Kagi LLM Benchmarkas gpt-5-6-luna | 49.1% | 14%of 22 | Kagi |
| SimpleBenchas gpt-5.6-luna_unknown | 46.8% | 10%of 30 | Epoch AI Benchmarking Hub |
| Reasoning & knowledge · Reasoning 47 | |||
| CritPt (physics research)as gpt-5.6-luna_max | 20.6% | 73%of 89 | Epoch AI Benchmarking Hub |
| GPQA Diamondas gpt-5.6-luna_max | 91.6% | 68%of 70 | Epoch AI Benchmarking Hub |
| Chess Puzzlesas gpt-5.6-luna_max | 43.0% | 63%of 73 | Epoch AI Benchmarking Hub |
| Mystery Game Puzzlesas gpt-5.6-luna_max | 21.0% | 39%of 57 | Epoch AI Benchmarking Hub |
| SimpleQA Verifiedas gpt-5.6-luna_max | 41.0% | 37%of 57 | Epoch AI Benchmarking Hub |
| ARC-AGI-2as gpt-5.6-luna_max | 59.5% | 28%of 44 | Epoch AI Benchmarking Hub |
| Blueprint-Bench 2as gpt-5.6-luna_unknown | 22.6% | 20%of 31 | Epoch AI Benchmarking Hub |
| Math · Math 72 | |||
| FrontierMath Tiers 1-3as gpt-5.6-luna_max | 82.1% | 76%of 60 | Epoch AI Benchmarking Hub |
| OTIS Mock AIMEas gpt-5.6-luna_max | 98.3% | 72%of 73 | Epoch AI Benchmarking Hub |
| FrontierMath Tier 4as gpt-5.6-luna_max | 61.0% | 71%of 52 | Epoch AI Benchmarking Hub |
| ProofBenchas gpt-5.6-luna_max | 60.0% | 70%of 68 | Epoch AI Benchmarking Hub |
| Coding · Coding 52 | |||
| ALE-Benchas gpt-5.6-luna_max | 1667.4% | 83%of 71 | Epoch AI Benchmarking Hub |
| DeepSWEas gpt-5.6-luna_max | 67.2% | 64%of 26 | Epoch AI Benchmarking Hub |
| SciCodeas gpt-5.6-luna_max | 53.6% | 61%of 86 | Epoch AI Benchmarking Hub |
| WeirdMLas gpt-5.6-luna_high | 60.9% | 51%of 52 | Epoch AI Benchmarking Hub |
| FrontierCodeas gpt-5.6-luna_unknown | 39.8% | 41%of 45 | Epoch AI Benchmarking Hub |
| CursorBenchas gpt-5.6-luna_max | 35.9% | 13%of 17 | Epoch AI Benchmarking Hub |
| Agents & real work · Agents 40 | |||
| LMCAas gpt-5.6-luna_max | 48.5% | 64%of 67 | Epoch AI Benchmarking Hub |
| GDP.pdfas gpt-5.6-luna_unknown | 22.7% | 44%of 37 | Epoch AI Benchmarking Hub |
| DTBenchas gpt-5.6-luna_xhigh | 89.1% | 39%of 67 | Epoch AI Benchmarking Hub |
| APEX-Agentsas gpt-5.6-luna_unknown | 43.0% | 27%of 48 | Epoch AI Benchmarking Hub |
| Vending-Bench 2as gpt-5.6-luna_unknown | 4094.7% | 26%of 47 | Epoch AI Benchmarking Hub |
| Independent indexes · Indexes 70 | |||
| Epoch Capabilities Indexas GPT-5.6 Luna | 156.3 | 70%of 81 | Epoch AI |