VISTA (Visually Impaired Scene and Task Annotations)
Models describe scenes, answer questions, and give guidance to blind or low-vision users from glasses recordings. A judge scores responses from 0 to 5.
The authors report scores for 14 vision-language models on 1,828 items from 450 test clips. Those models received eight camera frames and the task text.
- Made by
- Harvard AI and Robotics Lab and collaborators
- Released
- October 2026
- Kind
- Video understanding
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Vision
- Strongest model tested
-
Gemini 3.1 Flash-Lite #96 on the capability index
- Last model added
- March 2026
- Scores refreshed
- November 20th, 2026
Gemini 3.1 Flash-Lite leads with 4 among models released by March 2026 Gemini 3.1 Flash-Lite leads with 4 among models released by March 2026
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. 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.
The newest model with a score here came out in March 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
Bars start at 1.5.
- 1
Gemini 3.1 Flash-Lite Frames + text 4
- 2 GPT-Realtime mini Dec 2025, Frames + text 3
- 3
InternVL3-14B Frames + text 3
- 4 Qwen3-VL 8B Frames + text 3
- 5 Qwen3-VL 4B Frames + text 3
- 6
InternVL3-8B Frames + text 3
- 7
InternVL3-9B Frames + text 3
- 8
Gemma 3 12B Frames + text 3
- 9
Gemma 3 27B Frames + text 3
- 10
InternVL3-2B Frames + text 3
- 11
Gemma 3 4B Frames + text 3
- 12 Qwen3-VL 2B Frames + text 3
- 13
InternVL3-1B Frames + text 2
- 14 VideoLLAMA3-7B Frames + text 2
Source: Official VISTA (Visually Impaired Scene and Task Annotations) leaderboard
Bars start at 1.5.
- 1
Gemini 3.1 Flash-Lite Frames + text 4
- 2 GPT-Realtime mini Dec 2025, Frames + text 3
- 3
InternVL3-14B Frames + text 3
- 4
InternVL3-8B Frames + text 3
- 5
InternVL3-9B Frames + text 3
- 6
Gemma 3 12B Frames + text 3
- 7
Gemma 3 27B Frames + text 3
- 8
InternVL3-2B Frames + text 3
- 9
Gemma 3 4B Frames + text 3
- 10
InternVL3-1B Frames + text 2
- 11 VideoLLAMA3-7B Frames + text 2
Source: Official VISTA (Visually Impaired Scene and Task Annotations) leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 4 | Frames + text | Mar 2026 | Official leaderboard |
| 2 | GPT-Realtime mini Dec 2025 | 3 | Frames + text | Dec 2025 | Official leaderboard |
| 3 | | 3 | Frames + text | Apr 2025 | Official leaderboard |
| 4 | Qwen3-VL 8B open weights, 8B | 3 | Frames + text | Official leaderboard | |
| 5 | Qwen3-VL 4B open weights, 4B | 3 | Frames + text | Official leaderboard | |
| 6 | | 3 | Frames + text | Apr 2025 | Official leaderboard |
| 7 | | 3 | Frames + text | Apr 2025 | Official leaderboard |
| 8 | | 3 | Frames + text | Mar 2025 | Official leaderboard |
| 9 | | 3 | Frames + text | Mar 2025 | Official leaderboard |
| 10 | | 3 | Frames + text | Apr 2025 | Official leaderboard |
| 11 | | 3 | Frames + text | Mar 2025 | Official leaderboard |
| 12 | Qwen3-VL 2B open weights, 2B | 3 | Frames + text | Official leaderboard | |
| 13 | | 2 | Frames + text | Apr 2025 | Official leaderboard |
| 14 | VideoLLAMA3-7B | 2 | Frames + text | Jan 2024 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 4 | Frames + text | Mar 2026 | Official leaderboard |
| 2 | GPT-Realtime mini Dec 2025 | 3 | Frames + text | Dec 2025 | Official leaderboard |
| 3 | | 3 | Frames + text | Apr 2025 | Official leaderboard |
| 4 | | 3 | Frames + text | Apr 2025 | Official leaderboard |
| 5 | | 3 | Frames + text | Apr 2025 | Official leaderboard |
| 6 | | 3 | Frames + text | Mar 2025 | Official leaderboard |
| 7 | | 3 | Frames + text | Mar 2025 | Official leaderboard |
| 8 | | 3 | Frames + text | Apr 2025 | Official leaderboard |
| 9 | | 3 | Frames + text | Mar 2025 | Official leaderboard |
| 10 | | 2 | Frames + text | Apr 2025 | Official leaderboard |
| 11 | VideoLLAMA3-7B | 2 | Frames + text | Jan 2024 | Official leaderboard |
Rankings by part
We read 3 parts from the benchmark’s own leaderboard: Scene, Q&A and Guidance. Each ranks models on one part of the test. Gemini 3.1 Flash-Lite leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.
Scene
11 models we track have a score here.
- 1
Gemini 3.1 Flash-Lite 4
- 2 GPT-Realtime mini Dec 2025 4
- 3
InternVL3-14B 4
- 4
InternVL3-2B ↑4 4
- 5
InternVL3-8B ↓1 4
- 6
InternVL3-9B ↓1 4
- 7
InternVL3-1B ↑3 3
- 8
Gemma 3 12B ↓2 3
- 9
Gemma 3 27B ↓2 3
- 10
Gemma 3 4B ↓1 3
Q&A
11 models we track have a score here.
- 1
Gemini 3.1 Flash-Lite 4
- 2 GPT-Realtime mini Dec 2025 4
- 3
Gemma 3 27B ↑4 4
- 4
Gemma 3 12B ↑2 4
- 5
InternVL3-14B ↓2 4
- 6
Gemma 3 4B ↑3 3
- 7
InternVL3-8B ↓3 3
- 8
InternVL3-9B ↓3 3
- 9
InternVL3-2B ↓1 3
- 10
InternVL3-1B 3
Guidance
11 models we track have a score here.
- 1
Gemini 3.1 Flash-Lite 3
- 2
InternVL3-14B ↑1 3
- 3 GPT-Realtime mini Dec 2025 ↓1 3
- 4
InternVL3-9B ↑1 3
- 5
InternVL3-8B ↓1 3
- 6
Gemma 3 27B ↑1 2
- 7
Gemma 3 12B ↓1 2
- 8
Gemma 3 4B ↑1 2
- 9
InternVL3-2B ↓1 2
- 10
InternVL3-1B 1
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official VISTA (Visually Impaired Scene and Task Annotations) leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official VISTA (Visually Impaired Scene and Task Annotations) 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 Audio and video
- Video-MMMU In use 300 expert-level lecture videos with questions testing perception, comprehension and adaptation of knowledge. Accuracy.
Qwen 3.8 Max 88.7%
- LVBench In use Hours-long video understanding with multiple-choice questions on reasoning, grounding and summarization.
Gemini 4 Argon 91.7%
- MMVU In use 3,000 expert-annotated questions on videos across 27 disciplines. Multiple-choice and open-ended accuracy.
Qwen 3.8 Max 82.4%
- MLVU In use Multi-task long video understanding (minutes to hours) with multiple-choice and generation tasks. Accuracy.
Qwen 3.8 Max 90.8%
- FLEURS In use Models transcribe read speech in 102 languages built from FLoRes-101 sentences. Model cards usually report word error rate, so lower scores are better.
Phi-4-multimodal 4.00%
- VoiceBench In use Spoken instructions with speaker, environment and content variation. Voice assistant responses scored by GPT judge or rules.
Qwen3.5-Omni-Plus 93.1%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 8 months ago, in March 2026, for Gemini 3.1 Flash-Lite. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (14 entries).
- Pass Models scored 11 models we track, plus 3 other entries on its leaderboard.
- Warning Newest score The newest came 8 months ago, in March 2026, for Gemini 3.1 Flash-Lite.
- Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3.1 Flash-Lite (#96).
- Note Headroom Scored as a rating, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on harvard-ai-and-robotics-lab.github.io: 11 rows for models we track and 3 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.