ANCHOR
Models explain answers to audiovisual questions and locate the supporting evidence. The benchmark includes time intervals, image regions, and relevant sounds.
- Made by
- Joel Julin and colleagues
- Released
- October 2026
- Kind
- Audio understanding and voice
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Vision, Audio and speech
- Strongest model tested
- GPT-5.2 #51 on the capability index
- Last model added
- April 2026
- Scores refreshed
- November 20th, 2026
Gemini 3 Flash leads with 3.36 among models released by April 2026 Gemini 3 Flash leads with 3.36 among models released by April 2026
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The authors report four separate scores from 1 to 5. The headline is their plain mean on the main evaluation. 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 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
- 1
Gemini 3 Flash Truncated-video input 3.36
- 2 Qwen3-omni Truncated-video input 2.98
- 3
Nemotron 3 Nano Omni 30B-A3B Truncated-video input 2.96
- 4 GPT-5.2 Truncated-video input 2.84
- 5
Qwen3-VL-30B-A3B-Instruct Truncated-video input 2.79
- 6 GLM-4.6V-Flash Truncated-video input 2.55
- 7 Kimi-VL-A3B-Instruct Truncated-video input 2.53
- 8
Amazon Nova Lite Truncated-video input 2.34
- 9 Llava-OneVision Truncated-video input 2.30
- 10 Claude Sonnet 4.6 Truncated-video input 2.27
- 11 InternVL3.5 Truncated-video input 2.25
- 12 Grok 4.1 Fast Truncated-video input 2.14
Source: Official ANCHOR leaderboard
- 1
Gemini 3 Flash Truncated-video input 3.36
- 2
Nemotron 3 Nano Omni 30B-A3B Truncated-video input 2.96
- 3 GPT-5.2 Truncated-video input 2.84
- 4
Qwen3-VL-30B-A3B-Instruct Truncated-video input 2.79
- 5 GLM-4.6V-Flash Truncated-video input 2.55
- 6 Kimi-VL-A3B-Instruct Truncated-video input 2.53
- 7
Amazon Nova Lite Truncated-video input 2.34
- 8 Claude Sonnet 4.6 Truncated-video input 2.27
- 9 Grok 4.1 Fast Truncated-video input 2.14
Source: Official ANCHOR leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 3.36 | Truncated-video input | Dec 2025 | Official leaderboard |
| 2 | Qwen3-omni | 2.98 | Truncated-video input | Official leaderboard | |
| 3 | | 2.96 | Truncated-video input | Apr 2026 | Official leaderboard |
| 4 | GPT-5.2 | 2.84 | Truncated-video input | Dec 2025 | Official leaderboard |
| 5 | | 2.79 | Truncated-video input | Oct 2025 | Official leaderboard |
| 6 | GLM-4.6V-Flash | 2.55 | Truncated-video input | Dec 2025 | Official leaderboard |
| 7 | Kimi-VL-A3B-Instruct | 2.53 | Truncated-video input | Apr 2025 | Official leaderboard |
| 8 | | 2.34 | Truncated-video input | Dec 2024 | Official leaderboard |
| 9 | Llava-OneVision | 2.30 | Truncated-video input | Official leaderboard | |
| 10 | Claude Sonnet 4.6 | 2.27 | Truncated-video input | Feb 2026 | Official leaderboard |
| 11 | InternVL3.5 | 2.25 | Truncated-video input | Official leaderboard | |
| 12 | Grok 4.1 Fast | 2.14 | Truncated-video input | Nov 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 3.36 | Truncated-video input | Dec 2025 | Official leaderboard |
| 2 | | 2.96 | Truncated-video input | Apr 2026 | Official leaderboard |
| 3 | GPT-5.2 | 2.84 | Truncated-video input | Dec 2025 | Official leaderboard |
| 4 | | 2.79 | Truncated-video input | Oct 2025 | Official leaderboard |
| 5 | GLM-4.6V-Flash | 2.55 | Truncated-video input | Dec 2025 | Official leaderboard |
| 6 | Kimi-VL-A3B-Instruct | 2.53 | Truncated-video input | Apr 2025 | Official leaderboard |
| 7 | | 2.34 | Truncated-video input | Dec 2024 | Official leaderboard |
| 8 | Claude Sonnet 4.6 | 2.27 | Truncated-video input | Feb 2026 | Official leaderboard |
| 9 | Grok 4.1 Fast | 2.14 | Truncated-video input | Nov 2025 | Official leaderboard |
Rankings by part
We read 4 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Answer, Hallucination and Divergence. Gemini 3 Flash leads 3 of them, and Nemotron 3 Nano Omni 30B-A3B leads the other one. Arrows show how far a model moved from its overall rank among the models we track.
Answer
9 models we track have a score here.
- 1
Gemini 3 Flash 3.83
- 2 GPT-5.2 ↑1 3.77
- 3
Qwen3-VL-30B-A3B-Instruct ↑1 3.45
- 4 Grok 4.1 Fast ↑5 3.39
- 5 Claude Sonnet 4.6 ↑3 3.37
- 6
Nemotron 3 Nano Omni 30B-A3B ↓4 3.37
- 7
Amazon Nova Lite 3.30
- 8 GLM-4.6V-Flash ↓3 3.29
- 9 Kimi-VL-A3B-Instruct ↓3 3.05
Hallucination
9 models we track have a score here.
- 1
Nemotron 3 Nano Omni 30B-A3B ↑1 2.21
- 2
Gemini 3 Flash ↓1 2.20
- 3 GPT-5.2 2.05
- 4 GLM-4.6V-Flash ↑1 1.94
- 5 Kimi-VL-A3B-Instruct ↑1 1.93
- 6
Qwen3-VL-30B-A3B-Instruct ↓2 1.79
- 7 Claude Sonnet 4.6 ↑1 1.66
- 8
Amazon Nova Lite ↓1 1.63
- 9 Grok 4.1 Fast 1.48
Divergence
9 models we track have a score here.
- 1
Gemini 3 Flash 2.93
- 2 GPT-5.2 ↑1 2.83
- 3
Nemotron 3 Nano Omni 30B-A3B ↓1 2.59
- 4 Claude Sonnet 4.6 ↑4 2.55
- 5
Qwen3-VL-30B-A3B-Instruct ↓1 2.50
- 6 GLM-4.6V-Flash ↓1 2.46
- 7
Amazon Nova Lite 2.42
- 8 Grok 4.1 Fast ↑1 2.39
- 9 Kimi-VL-A3B-Instruct ↓3 2.31
Conciseness
9 models we track have a score here.
- 1
Gemini 3 Flash 4.47
- 2
Nemotron 3 Nano Omni 30B-A3B 3.69
- 3
Qwen3-VL-30B-A3B-Instruct ↑1 3.40
- 4 Kimi-VL-A3B-Instruct ↑2 2.83
- 5 GPT-5.2 ↓2 2.72
- 6 GLM-4.6V-Flash ↓1 2.50
- 7
Amazon Nova Lite 2.01
- 8 Claude Sonnet 4.6 1.50
- 9 Grok 4.1 Fast 1.30
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official ANCHOR 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 ANCHOR 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 7 months ago, in April 2026, for Nemotron 3 Nano Omni 30B-A3B. 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 (12 entries).
- Pass Models scored 9 models we track, plus 3 other entries on its leaderboard.
- Warning Newest score The newest came 7 months ago, in April 2026, for Nemotron 3 Nano Omni 30B-A3B.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.2 (#51).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on anchorbenchmark.github.io: 9 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.