MVDriveBench
Models answer driving questions using synchronized videos from cameras facing different directions. The authors compare surround-view results with front-camera results.
- Made by
- Shunming Liu and colleagues
- Released
- November 2026
- Kind
- Embodied reasoning
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Multiple choice
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- A host
- In views
- Vision
- Strongest model tested
- GPT-5.5 #21 on the capability index
- Last model added
- April 2026
- Scores refreshed
- November 20th, 2026
GPT-5.5 leads with 80.1% among models released by April 2026 GPT-5.5 leads with 80.1% among models released by April 2026
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 80.1 GPT-5.5
- 78.7 GPT-5
- 78.4
Gemini 3 Pro
- 76.7 Gemini-2.5-Pro (snapshot unspecified)
- 75.9 Claude Opus 4.6
- 74.7
Qwen3.5 397B-A17B
- 74.3 GPT-4o (snapshot unspecified)
- 71.5
Qwen2.5-VL-32B
- 71.2
InternVL3-78B
- 70.9
InternVL3-38B
- 69.4
Qwen2.5-VL-72B
- 66.4
Qwen3-VL-30B-A3B-Thinking
- 62.2 Video-R1-7B
- 61.7
Qwen3.5-27B
- 56.4 VGLM-QA-VGGT-8B
- 54.4 MiMo-VL-7B-SFT
- 51.3 Cosmos-Reason2-8B
- 50.5
LLaVA-OneVision 7B
- 49.9
ERNIE-4.5-VL-28B-A3B-Thinking
- 47.3 MiMoEmbodied-7B
- 47.3 ReCogDrive-VLM-8B
- 47.2 GLM-4.1V-9B-Thinking
- 45.7 Cosmos-Reason1-7B
- 1 GPT-5.5 Surround-view 80.1%
- 2 GPT-5 Surround-view 78.7%
- 3
Gemini 3 Pro Front-view 78.4%
- 4 Gemini-2.5-Pro (snapshot unspecified) Surround-view 76.7%
- 5 Claude Opus 4.6 Surround-view 75.9%
- 6
Qwen3.5 397B-A17B Surround-view 74.7%
- 7 GPT-4o (snapshot unspecified) Surround-view 74.3%
- 8
Qwen2.5-VL-32B Surround-view 71.5%
- 9
InternVL3-78B Surround-view 71.2%
- 10
InternVL3-38B Surround-view 70.9%
- 11
Qwen2.5-VL-72B Surround-view 69.4%
- 12
Qwen3-VL-30B-A3B-Thinking Surround-view 66.4%
- 13 Video-R1-7B Surround-view 62.2%
- 14
Qwen3.5-27B Surround-view 61.7%
- 15 VGLM-QA-VGGT-8B Surround-view 56.4%
- 16 MiMo-VL-7B-SFT Surround-view 54.4%
- 17 Cosmos-Reason2-8B Surround-view 51.3%
- 18
LLaVA-OneVision 7B Surround-view 50.5%
- 19
ERNIE-4.5-VL-28B-A3B-Thinking Surround-view 49.9%
- 20 MiMoEmbodied-7B Surround-view 47.3%
- 21 ReCogDrive-VLM-8B Surround-view 47.3%
- 22 GLM-4.1V-9B-Thinking Front-view 47.2%
- 23 Cosmos-Reason1-7B Surround-view 45.7%
Source: The MVDriveBench paper
- 80.1 GPT-5.5
- 78.7 GPT-5
- 78.4
Gemini 3 Pro
- 75.9 Claude Opus 4.6
- 74.7
Qwen3.5 397B-A17B
- 71.5
Qwen2.5-VL-32B
- 71.2
InternVL3-78B
- 70.9
InternVL3-38B
- 69.4
Qwen2.5-VL-72B
- 66.4
Qwen3-VL-30B-A3B-Thinking
- 61.7
Qwen3.5-27B
- 54.4 MiMo-VL-7B-SFT
- 50.5
LLaVA-OneVision 7B
- 49.9
ERNIE-4.5-VL-28B-A3B-Thinking
- 47.2 GLM-4.1V-9B-Thinking
- 1 GPT-5.5 Surround-view 80.1%
- 2 GPT-5 Surround-view 78.7%
- 3
Gemini 3 Pro Front-view 78.4%
- 4 Claude Opus 4.6 Surround-view 75.9%
- 5
Qwen3.5 397B-A17B Surround-view 74.7%
- 6
Qwen2.5-VL-32B Surround-view 71.5%
- 7
InternVL3-78B Surround-view 71.2%
- 8
InternVL3-38B Surround-view 70.9%
- 9
Qwen2.5-VL-72B Surround-view 69.4%
- 10
Qwen3-VL-30B-A3B-Thinking Surround-view 66.4%
- 11
Qwen3.5-27B Surround-view 61.7%
- 12 MiMo-VL-7B-SFT Surround-view 54.4%
- 13
LLaVA-OneVision 7B Surround-view 50.5%
- 14
ERNIE-4.5-VL-28B-A3B-Thinking Surround-view 49.9%
- 15 GLM-4.1V-9B-Thinking Front-view 47.2%
Source: The MVDriveBench paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-5.5 | 80.1% | Surround-view | Apr 2026 | Official leaderboard |
| 2 | GPT-5 | 78.7% | Surround-view | Aug 2025 | Official leaderboard |
| 3 | | 78.4% | Front-view | Nov 2025 | Official leaderboard |
| 4 | Gemini-2.5-Pro (snapshot unspecified) | 76.7% | Surround-view | Official leaderboard | |
| 5 | Claude Opus 4.6 | 75.9% | Surround-view | Feb 2026 | Official leaderboard |
| 6 | | 74.7% | Surround-view | Feb 2026 | Official leaderboard |
| 7 | GPT-4o (snapshot unspecified) | 74.3% | Surround-view | Official leaderboard | |
| 8 | | 71.5% | Surround-view | Mar 2025 | Official leaderboard |
| 9 | | 71.2% | Surround-view | Apr 2025 | Official leaderboard |
| 10 | | 70.9% | Surround-view | Apr 2025 | Official leaderboard |
| 11 | | 69.4% | Surround-view | Jan 2025 | Official leaderboard |
| 12 | | 66.4% | Surround-view | Oct 2025 | Official leaderboard |
| 13 | Video-R1-7B | 62.2% | Surround-view | Official leaderboard | |
| 14 | | 61.7% | Surround-view | Feb 2026 | Official leaderboard |
| 15 | VGLM-QA-VGGT-8B | 56.4% | Surround-view | Official leaderboard | |
| 16 | MiMo-VL-7B-SFT | 54.4% | Surround-view | May 2025 | Official leaderboard |
| 17 | Cosmos-Reason2-8B | 51.3% | Surround-view | Official leaderboard | |
| 18 | | 50.5% | Surround-view | Aug 2024 | Official leaderboard |
| 19 | | 49.9% | Surround-view | Nov 2025 | Official leaderboard |
| 20 | MiMoEmbodied-7B | 47.3% | Surround-view | Official leaderboard | |
| 21 | ReCogDrive-VLM-8B | 47.3% | Surround-view | Official leaderboard | |
| 22 | GLM-4.1V-9B-Thinking | 47.2% | Front-view | Jul 2025 | Official leaderboard |
| 23 | Cosmos-Reason1-7B | 45.7% | Surround-view | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-5.5 | 80.1% | Surround-view | Apr 2026 | Official leaderboard |
| 2 | GPT-5 | 78.7% | Surround-view | Aug 2025 | Official leaderboard |
| 3 | | 78.4% | Front-view | Nov 2025 | Official leaderboard |
| 4 | Claude Opus 4.6 | 75.9% | Surround-view | Feb 2026 | Official leaderboard |
| 5 | | 74.7% | Surround-view | Feb 2026 | Official leaderboard |
| 6 | | 71.5% | Surround-view | Mar 2025 | Official leaderboard |
| 7 | | 71.2% | Surround-view | Apr 2025 | Official leaderboard |
| 8 | | 70.9% | Surround-view | Apr 2025 | Official leaderboard |
| 9 | | 69.4% | Surround-view | Jan 2025 | Official leaderboard |
| 10 | | 66.4% | Surround-view | Oct 2025 | Official leaderboard |
| 11 | | 61.7% | Surround-view | Feb 2026 | Official leaderboard |
| 12 | MiMo-VL-7B-SFT | 54.4% | Surround-view | May 2025 | Official leaderboard |
| 13 | | 50.5% | Surround-view | Aug 2024 | Official leaderboard |
| 14 | | 49.9% | Surround-view | Nov 2025 | Official leaderboard |
| 15 | GLM-4.1V-9B-Thinking | 47.2% | Front-view | Jul 2025 | Official leaderboard |
Rankings by part
We read 15 parts from the benchmark’s paper, each ranking models on one part of the test, such as Rel. Mot., Obj. Mot. and Tracking. GPT-5.5 leads 5 of them, and 4 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Rel. Mot.
15 models we track have a score here.
- 1 GPT-5 ↑1 64.9%
- 2
Qwen3.5-27B ↑9 63.2%
- 3 GPT-5.5 ↓2 62.3%
- 4
Gemini 3 Pro ↓1 61.3%
- 5 Claude Opus 4.6 ↓1 60.9%
- 6
Qwen3.5 397B-A17B ↓1 57.9%
- 7
InternVL3-38B ↑1 53.2%
- 8
Qwen2.5-VL-32B ↓2 49.9%
- 9
Qwen3-VL-30B-A3B-Thinking ↑1 49%
- 10
InternVL3-78B ↓3 48.7%
Obj. Mot.
15 models we track have a score here.
- 1 GPT-5.5 65.4%
- 2 Claude Opus 4.6 ↑2 63.8%
- 3 GPT-5 ↓1 63.5%
- 4
Qwen3.5-27B ↑7 61.4%
- 5
InternVL3-78B ↑2 60%
- 6
Qwen3.5 397B-A17B ↓1 60%
- 7
Qwen3-VL-30B-A3B-Thinking ↑3 58.9%
- 8
InternVL3-38B 56%
- 9
Gemini 3 Pro ↓6 55.2%
- 10
Qwen2.5-VL-72B ↓1 49.1%
Tracking
15 models we track have a score here.
- 1
Gemini 3 Pro ↑2 65.2%
- 2 GPT-5 62.7%
- 3 GPT-5.5 ↓2 61.2%
- 4
Qwen3.5-27B ↑7 59.3%
- 5
Qwen3-VL-30B-A3B-Thinking ↑5 57.7%
- 6 Claude Opus 4.6 ↓2 57.3%
- 7
Qwen3.5 397B-A17B ↓2 54.1%
- 8
InternVL3-38B 50.3%
- 9
InternVL3-78B ↓2 49.2%
- 10
Qwen2.5-VL-72B ↓1 42.3%
Ego Mot.
15 models we track have a score here.
- 1
Gemini 3 Pro ↑2 89.6%
- 2
Qwen3.5 397B-A17B ↑3 89%
- 3 GPT-5 ↓1 87.9%
- 4 GPT-5.5 ↓3 86.8%
- 5
Qwen2.5-VL-32B ↑1 79.1%
- 6
InternVL3-38B ↑2 79.1%
- 7
InternVL3-78B 78.2%
- 8 Claude Opus 4.6 ↓4 76.6%
- 9 MiMo-VL-7B-SFT ↑3 76.3%
- 10
LLaVA-OneVision 7B ↑3 76%
Dynamic Motion Understanding
15 models we track have a score here.
- 1 GPT-5.5 68.9%
- 2 GPT-5 67.7%
- 3
Gemini 3 Pro 66.1%
- 4
Qwen3.5 397B-A17B ↑1 65.2%
- 5 Claude Opus 4.6 ↓1 64.7%
- 6
InternVL3-38B ↑2 59.6%
- 7
InternVL3-78B 59%
- 8
Qwen3.5-27B ↑3 59%
- 9
Qwen3-VL-30B-A3B-Thinking ↑1 54.4%
- 10
Qwen2.5-VL-32B ↓4 53.7%
Intent
15 models we track have a score here.
- 1 GPT-5.5 85.1%
- 2
Gemini 3 Pro ↑1 83.4%
- 3 GPT-5 ↓1 83%
- 4 MiMo-VL-7B-SFT ↑8 82.7%
- 5
Qwen2.5-VL-72B ↑4 82.2%
- 6
Qwen3-VL-30B-A3B-Thinking ↑4 80.7%
- 7
Qwen3.5 397B-A17B ↓2 78.9%
- 8 Claude Opus 4.6 ↓4 76.6%
- 9
Qwen2.5-VL-32B ↓3 76%
- 10
InternVL3-38B ↓2 71.3%
Cross-view
15 models we track have a score here.
- 1
InternVL3-78B ↑6 91.1%
- 2
Gemini 3 Pro ↑1 87.3%
- 3
InternVL3-38B ↑5 86.2%
- 4
Qwen2.5-VL-32B ↑2 83.8%
- 5 MiMo-VL-7B-SFT ↑7 83.5%
- 6 Claude Opus 4.6 ↓2 80.9%
- 7 GPT-5.5 ↓6 76.7%
- 8 GPT-5 ↓6 75.7%
- 9
Qwen3.5 397B-A17B ↓4 70%
- 10
Qwen2.5-VL-72B ↓1 63.5%
Key Obj.
15 models we track have a score here.
- 1
Qwen3.5-27B ↑10 81.6%
- 2 Claude Opus 4.6 ↑2 75.8%
- 3
Gemini 3 Pro 74%
- 4 GPT-5.5 ↓3 73.5%
- 5 GPT-5 ↓3 72.3%
- 6
Qwen3-VL-30B-A3B-Thinking ↑4 70.6%
- 7
Qwen3.5 397B-A17B ↓2 69%
- 8
Qwen2.5-VL-32B ↓2 63.8%
- 9
InternVL3-78B ↓2 62.5%
- 10
Qwen2.5-VL-72B ↓1 61.3%
Key Event
15 models we track have a score here.
- 1
Qwen3.5-27B ↑10 80.7%
- 2 GPT-5 78.1%
- 3 GPT-5.5 ↓2 76.9%
- 4
Gemini 3 Pro ↓1 75.6%
- 5
Qwen3-VL-30B-A3B-Thinking ↑5 70.1%
- 6 Claude Opus 4.6 ↓2 70%
- 7
Qwen3.5 397B-A17B ↓2 66.5%
- 8
Qwen2.5-VL-72B ↑1 62.6%
- 9
InternVL3-78B ↓2 62.3%
- 10
InternVL3-38B ↓2 61.1%
Risk
15 models we track have a score here.
- 1 GPT-5 ↑1 96.6%
- 2
Gemini 3 Pro ↑1 95.1%
- 3 GPT-5.5 ↓2 92%
- 4
Qwen2.5-VL-32B ↑2 91.1%
- 5
Qwen3.5 397B-A17B 89.2%
- 6
Qwen2.5-VL-72B ↑3 88.9%
- 7 MiMo-VL-7B-SFT ↑5 88.4%
- 8
LLaVA-OneVision 7B ↑5 86.1%
- 9
InternVL3-38B ↓1 85.6%
- 10
InternVL3-78B ↓3 82.3%
Multi-view Scene Understanding
15 models we track have a score here.
- 1 GPT-5.5 80.8%
- 2
Gemini 3 Pro ↑1 80.1%
- 3 GPT-5 ↓1 79.3%
- 4 Claude Opus 4.6 77%
- 5
Qwen3.5 397B-A17B 74.6%
- 6
Qwen2.5-VL-32B 74.4%
- 7
InternVL3-78B 72.3%
- 8
InternVL3-38B 71.3%
- 9
Qwen2.5-VL-72B 71.2%
- 10
Qwen3-VL-30B-A3B-Thinking 63.5%
Planning
15 models we track have a score here.
- 1
Gemini 3 Pro ↑2 100%
- 2
InternVL3-38B ↑6 98.6%
- 3
Qwen2.5-VL-72B ↑6 98.3%
- 4
Qwen2.5-VL-32B ↑2 98.3%
- 5
InternVL3-78B ↑2 97.2%
- 6 GPT-5 ↓4 96.5%
- 7
Qwen3.5 397B-A17B ↓2 96.1%
- 8 MiMo-VL-7B-SFT ↑4 95%
- 9 GPT-5.5 ↓8 94%
- 10
Qwen3-VL-30B-A3B-Thinking 92.3%
Counterfactual
15 models we track have a score here.
- 1 GPT-5.5 95.1%
- 2
Gemini 3 Pro ↑1 94.8%
- 3 GPT-5 ↓1 93.4%
- 4
InternVL3-78B ↑3 91.5%
- 5
InternVL3-38B ↑3 90.7%
- 6
Qwen2.5-VL-72B ↑3 90.4%
- 7
Qwen2.5-VL-32B ↓1 90.4%
- 8 Claude Opus 4.6 ↓4 90.2%
- 9
Qwen3-VL-30B-A3B-Thinking ↑1 90.1%
- 10
Qwen3.5 397B-A17B ↓5 86.7%
Rationality
15 models we track have a score here.
- 1 GPT-5 ↑1 94.5%
- 2 GPT-5.5 ↓1 92.9%
- 3 Claude Opus 4.6 ↑1 91.5%
- 4
Qwen3.5-27B ↑7 89.9%
- 5
Gemini 3 Pro ↓2 87.8%
- 6
Qwen3-VL-30B-A3B-Thinking ↑4 81.6%
- 7
Qwen2.5-VL-32B ↓1 81.2%
- 8
Qwen3.5 397B-A17B ↓3 79.8%
- 9
Qwen2.5-VL-72B 79.1%
- 10 GLM-4.1V-9B-Thinking ↑5 77.4%
Driving Reasoning
15 models we track have a score here.
- 1
Gemini 3 Pro ↑2 94.2%
- 2 GPT-5.5 ↓1 94%
- 3 GPT-5 ↓1 93.9%
- 4
Qwen2.5-VL-32B ↑2 90%
- 5 Claude Opus 4.6 ↓1 89.4%
- 6
Qwen2.5-VL-72B ↑3 89.1%
- 7
Qwen3-VL-30B-A3B-Thinking ↑3 87.9%
- 8
Qwen3.5 397B-A17B ↓3 87.5%
- 9
InternVL3-78B ↓2 85.2%
- 10
InternVL3-38B ↓2 85%
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: The MVDriveBench paper
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| GPT-5.5 | Surround-view | 80.1% | |||
| GPT-5.5 | Front-view | 78.8% | |||
| GPT-5 | Surround-view | 78.7% | |||
| Front-view | 78.4% | ||||
| Surround-view | 78.3% | ||||
| GPT-5 | Front-view | 77.2% | |||
| Claude Opus 4.6 | Surround-view | 75.9% | |||
| Surround-view | 74.7% | ||||
| Claude Opus 4.6 | Front-view | 73.9% | |||
| Surround-view | 71.5% | ||||
| Surround-view | 71.2% | ||||
| Surround-view | 70.9% | ||||
| Surround-view | 69.4% | ||||
| Surround-view | 66.4% | ||||
| Surround-view | 61.7% | ||||
| Front-view | 59.6% | ||||
| Front-view | 56.2% | ||||
| Front-view | 55.9% | ||||
| Front-view | 55.6% | ||||
| MiMo-VL-7B-SFT | Surround-view | 54.4% | |||
| Front-view | 53.6% | ||||
| Front-view | 52.4% | ||||
| Surround-view | 50.5% | ||||
| Surround-view | 49.9% | ||||
| Front-view | 47.5% | ||||
| GLM-4.1V-9B-Thinking | Front-view | 47.2% | |||
| Front-view | 47.1% | ||||
| MiMo-VL-7B-SFT | Front-view | 44.9% | |||
| Front-view | 43.6% | ||||
| GLM-4.1V-9B-Thinking | Surround-view | 43.2% |
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: The MVDriveBench paper
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| GPT-5.5 | Surround-view | 80.1% | |||
| GPT-5.5 | Front-view | 78.8% | |||
| GPT-5 | Surround-view | 78.7% | |||
| Front-view | 78.4% | ||||
| Surround-view | 78.3% | ||||
| GPT-5 | Front-view | 77.2% | |||
| Claude Opus 4.6 | Surround-view | 75.9% | |||
| Surround-view | 74.7% | ||||
| Claude Opus 4.6 | Front-view | 73.9% | |||
| Surround-view | 71.5% | ||||
| Surround-view | 71.2% | ||||
| Surround-view | 70.9% | ||||
| Surround-view | 69.4% | ||||
| Surround-view | 66.4% | ||||
| Surround-view | 61.7% | ||||
| Front-view | 59.6% | ||||
| Front-view | 56.2% | ||||
| Front-view | 55.9% | ||||
| Front-view | 55.6% | ||||
| MiMo-VL-7B-SFT | Surround-view | 54.4% | |||
| Front-view | 53.6% | ||||
| Front-view | 52.4% | ||||
| Surround-view | 50.5% | ||||
| Surround-view | 49.9% | ||||
| Front-view | 47.5% | ||||
| GLM-4.1V-9B-Thinking | Front-view | 47.2% | |||
| Front-view | 47.1% | ||||
| MiMo-VL-7B-SFT | Front-view | 44.9% | |||
| Front-view | 43.6% | ||||
| GLM-4.1V-9B-Thinking | Surround-view | 43.2% |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The MVDriveBench paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official MVDriveBench leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Robotics
- ERQA In use 400 multiple-choice visual questions from robot and egocentric footage on spatial, trajectory, action and state reasoning. Accuracy.
Gemini Robotics ER 2 78.5%
- SimplerEnv In use Real-to-sim replicas of Google Robot and WidowX/Bridge setups. Policies scored by success rate on pick, move, drawer and Bridge tasks. Praxis-1 80%
- RoboLab In use Zero-shot DROID Franka policies in Isaac Lab on 120 tabletop tasks at three instruction levels, 10 episodes each. Success rate. Praxis-1 46.2%
- RoboDojo In use Generalist manipulation policies run 42 Isaac Sim tasks and 18 remote real-robot tasks. Scored by task score and success rate. GPT-6 Astra 27.6%
- Progress classification (Google DeepMind) In use Models assign each frame of a robot video to one of five task-progress bands from 0 to 100 percent. Scored by accuracy.
Gemini Robotics ER 2 57.4%
- Human proximity (Google DeepMind robotics) In use Models judge from robot camera views whether a human is within 1 meter, a spatial awareness check for safe robot behavior. Scored by accuracy.
Gemini Robotics ER 2 93%
Data audit
8 of 10 checks pass, 2 warnings. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark
Catalog record 1 warning 1 pass
- Warning Task access We don’t know whether the tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the benchmark’s paper (23 entries).
- Pass Models scored 15 models we track, plus 8 other entries on its leaderboard.
- Warning Newest score The newest came 7 months ago, in April 2026, for GPT-5.5.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
- Pass Headroom The best is 80.1%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 30 rows for models we track and 8 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.