Benchmark Almanac

RSLiveBench-UAV

GPT-6 Astra 59.9 leads since November 2026
Not in model cards

Models answer questions about vehicles in drone images. The benchmark checks perception and spatial reasoning across different views and adds newly captured scenes.

Released
November 2026
Kind
Embodied reasoning
Level
Everyday
Scored
Checked automatically
Answers
Short answer
Tasks
Public, plus a private test set
Measures
Capability
Language
English
In and out
Images, Text → Text
Interaction
Single turn
Items from
Real work and use
Run by
A host
In views
Vision, Kept fresh
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
October 2026
Scores refreshed
November 20th, 2026
30
Entries in its paper, 10 models we track

GPT-6 Astra leads with 59.9 GPT-6 Astra leads with 59.9

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. 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.

  1. 1 GPT-6 Astra Low reasoning effort, Preliminary 59.9
  2. 2 GPT-6.1 Sol Low reasoning effort, Preliminary 59.1
  3. 3 ACE-Brain-0.6-9B Preliminary 43.3
  4. 4 GPT-5.6 Luna Low reasoning effort, Preliminary 37.8
  5. 5 DeepSeek V4.1 Flash 1017 Low reasoning effort, Preliminary 37.2
  6. 6 Qwen3.5-9B Preliminary 36.1
  7. 7 RynnBrain1.1-9B Preliminary 35.1
  8. 8 RoboBrain2.5-8B-NV Preliminary 32.8
  9. 9 RynnBrain-8B Preliminary 31.7
  10. 10 Qwen3-VL-8B-Instruct Preliminary 31.2
  11. 11 PhysBrain1.5-8B Preliminary 24.1
  12. 12 MiMo-Embodied-7B Preliminary 24.0
  13. 13 ACE-Brain-0-8B Preliminary 21.1
  14. 14 Cosmos-Reason2-8B Preliminary 20.8
  15. 15 SenseNova-SI-1.4-InternVL3-8B Preliminary 18.9
  16. 16 InternVL3.5-4B Preliminary 18.8
  17. 17 InternVL3.5-8B Preliminary 17.6
  18. 18 SenseNova-SI-1.5-InternVL3-8B Preliminary 16.9
  19. 19 SenseNova-SI-1.3-InternVL3-8B Preliminary 16.6
  20. 20 Qwen3.5-2B Preliminary 15.2
  21. 21 ACE-Brain-0.5-8B Preliminary 14.9
  22. 22 SenseNova-SI-1.1-InternVL3-8B Preliminary 14.8
  23. 23 RynnBrain1.1-2B Preliminary 14.5
  24. 24 Cosmos-Reason2-2B Preliminary 14.3
  25. 25 RynnBrain-2B Preliminary 14.1
  26. 26 Pelican1.0-VL-7B 2K, Preliminary 11.9
  27. 27 SenseNova-SI-1.1-InternVL3-2B Preliminary 9.13
  28. 28 GeoChat-7B Preliminary 8.31
  29. 29 VHM Preliminary 4.38
  30. 30 GeoLLaVA-8K Preliminary 0.11

Source: The RSLiveBench-UAV paper

#EntryScoreSettingDateLinkSource
1 GPT-6 Astra 59.9 Low reasoning effort, Preliminary Sep 2026 Official leaderboard
2 GPT-6.1 Sol 59.1 Low reasoning effort, Preliminary Sep 2026 Official leaderboard
3 ACE-Brain-0.6-9B open weights, 9B 43.3 Preliminary huggingface.co Official leaderboard
4 GPT-5.6 Luna 37.8 Low reasoning effort, Preliminary Jul 2026 Official leaderboard
5 DeepSeek V4.1 Flash 1017 37.2 Low reasoning effort, Preliminary Oct 2026 deepseek.com Official leaderboard
6 Qwen3.5-9B 36.1 Preliminary Feb 2026 huggingface.co Official leaderboard
7 RynnBrain1.1-9B open weights, 9B 35.1 Preliminary huggingface.co Official leaderboard
8 RoboBrain2.5-8B-NV open weights, 8B 32.8 Preliminary huggingface.co Official leaderboard
9 RynnBrain-8B open weights, 8B 31.7 Preliminary huggingface.co Official leaderboard
10 Qwen3-VL-8B-Instruct 31.2 Preliminary Oct 2025 huggingface.co Official leaderboard
11 PhysBrain1.5-8B open weights, 8B 24.1 Preliminary deepcybo-physai.github.io Official leaderboard
12 MiMo-Embodied-7B open weights, 7B 24.0 Preliminary github.com Official leaderboard
13 ACE-Brain-0-8B open weights, 8B 21.1 Preliminary huggingface.co Official leaderboard
14 Cosmos-Reason2-8B open weights, 8B 20.8 Preliminary docs.nvidia.com Official leaderboard
15 SenseNova-SI-1.4-InternVL3-8B open weights, 8B 18.9 Preliminary huggingface.co Official leaderboard
16 InternVL3.5-4B 18.8 Preliminary Aug 2025 huggingface.co Official leaderboard
17 InternVL3.5-8B 17.6 Preliminary Aug 2025 huggingface.co Official leaderboard
18 SenseNova-SI-1.5-InternVL3-8B open weights, 8B 16.9 Preliminary huggingface.co Official leaderboard
19 SenseNova-SI-1.3-InternVL3-8B 16.6 Preliminary Jan 2026 huggingface.co Official leaderboard
20 Qwen3.5-2B 15.2 Preliminary Feb 2026 huggingface.co Official leaderboard
21 ACE-Brain-0.5-8B open weights, 8B 14.9 Preliminary huggingface.co Official leaderboard
22 SenseNova-SI-1.1-InternVL3-8B open weights, 8B 14.8 Preliminary huggingface.co Official leaderboard
23 RynnBrain1.1-2B open weights, 2B 14.5 Preliminary huggingface.co Official leaderboard
24 Cosmos-Reason2-2B open weights, 2B 14.3 Preliminary docs.nvidia.com Official leaderboard
25 RynnBrain-2B open weights, 2B 14.1 Preliminary huggingface.co Official leaderboard
26 Pelican1.0-VL-7B open weights, 7B 11.9 2K, Preliminary github.com Official leaderboard
27 SenseNova-SI-1.1-InternVL3-2B open weights, 2B 9.13 Preliminary huggingface.co Official leaderboard
28 GeoChat-7B open weights, 7B 8.31 Preliminary arxiv.org Official leaderboard
29 VHM open weights, 7B 4.38 Preliminary github.com Official leaderboard
30 GeoLLaVA-8K open weights, 7B 0.11 Preliminary github.com Official leaderboard
#ModelScoreSettingDateLinkSource
1 GPT-6 Astra 59.9 Low reasoning effort, Preliminary Sep 2026 Official leaderboard
2 GPT-6.1 Sol 59.1 Low reasoning effort, Preliminary Sep 2026 Official leaderboard
3 GPT-5.6 Luna 37.8 Low reasoning effort, Preliminary Jul 2026 Official leaderboard
4 DeepSeek V4.1 Flash 1017 37.2 Low reasoning effort, Preliminary Oct 2026 deepseek.com Official leaderboard
5 Qwen3.5-9B 36.1 Preliminary Feb 2026 huggingface.co Official leaderboard
6 Qwen3-VL-8B-Instruct 31.2 Preliminary Oct 2025 huggingface.co Official leaderboard
7 InternVL3.5-4B 18.8 Preliminary Aug 2025 huggingface.co Official leaderboard
8 InternVL3.5-8B 17.6 Preliminary Aug 2025 huggingface.co Official leaderboard
9 SenseNova-SI-1.3-InternVL3-8B 16.6 Preliminary Jan 2026 huggingface.co Official leaderboard
10 Qwen3.5-2B 15.2 Preliminary Feb 2026 huggingface.co Official leaderboard

Rankings by part

We read 21 parts from the benchmark’s paper in 5 groups: suite, 2d perception, 3d perception, cross-view perception and reasoning. Each ranks models on one part of the test. GPT-6 Astra leads 12 of them, and 6 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

2D perception

10 models we track have a score here.

  1. 1 GPT-6.1 Sol ↑1 54.3%
  2. 2 GPT-6 Astra ↓1 54.2%
  3. 3 Qwen3-VL-8B-Instruct ↑3 42.2%
  4. 4 Qwen3.5-9B ↑1 40.2%
  5. 5 DeepSeek V4.1 Flash 1017 ↓1 36.3%
  6. 6 Qwen3.5-2B ↑4 33.8%
  7. 7 GPT-5.6 Luna ↓4 33.6%
  8. 8 InternVL3.5-8B 24.2%
  9. 9 SenseNova-SI-1.3-InternVL3-8B 20.5%
  10. 10 InternVL3.5-4B ↓3 20%

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Benchmark paper 1 leader

Source: The RSLiveBench-UAV paper

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
In the benchmark’s paper
10.0 20.0 30.0 40.0 50.0 60.0 2026 10.0 20.0 30.0 40.0 50.0 60.0 2026

Source: Official RSLiveBench-UAV 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

All 172 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

All 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access Public tasks, plus a private test set.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the benchmark’s paper (30 entries).
  • Pass Models scored 10 models we track, plus 20 other entries on its leaderboard.
  • Pass Newest score The newest is from October 2026, for DeepSeek V4.1 Flash 1017.
  • Note Frontier models Only 2 of the top 10 have a score. The strongest tested is GPT-6 Astra (#4).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on rslivebench.space: 10 rows for models we track and 20 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.

How we check