Benchmark Almanac

VAHBench

Not in model cards

Agents turn short drama briefs into finished videos with specified characters, dialogue, and durations. The evaluator checks completeness, quality, reliability, and accuracy of production reports.

Made by
Hongbo Zhang and colleagues
Released
November 2026
Kind
Video generation
Level
University
Scored
Graded by a model
Answers
Agent task
Measures
Capability
Subject
Arts
Language
English
In and out
Text → Actions, Video, Text
Interaction
Agentic
Environment
APIs and tools
Refresh
One fixed set
Run by
A host
Strongest model tested
Claude Opus 5 #11 on the capability index
Last model added
July 2026
Scores refreshed
November 20th, 2026
32
Entries in its paper, 1 model we track

Claude Opus 5 leads with 51.8 Claude Opus 5 leads with 51.8

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 on its own with several settings, we keep its best. Agents other companies built around a model get a pale shade of that model's lab color. Gray marks blends and entries not in our model list, such as fine-tunes and older models. None of these count toward leaders. Agents built around a model, such as Claude Code + GLM-5.2, are their own entries. We don't count them as the model's score. This view shows only models we track, each at its best.

Source: The VAHBench paper

Source: The VAHBench paper

#EntryScoreSettingDateSource
1 Claude Opus 5 51.8 Claude Code Jul 2026 Official leaderboard
2 Claude Code + GLM-5.2 49.3 Jun 2026 Official leaderboard
3 Claude Code + DeepSeek-V4-Flash 46.2 Apr 2026 Official leaderboard
4 gpt-5.6 44.1 Claude Code Official leaderboard
5 Hermes + Claude Opus 5 42.0 Jul 2026 Official leaderboard
6 DeepSeek Harness + Claude Opus 5 41.7 Jul 2026 Official leaderboard
7 Hermes + GLM-5.2 39.9 Jun 2026 Official leaderboard
8 Claude Code + OpenMontage + GLM-5.2 38.6 Jun 2026 Official leaderboard
9 OpenClaw + OpenMontage + Claude Opus 5 38.4 Jul 2026 Official leaderboard
10 Claude Code + OpenMontage + Claude Opus 5 37.8 Jul 2026 Official leaderboard
11 Hermes + DeepSeek-V4-Flash 31.3 Apr 2026 Official leaderboard
12 OpenClaw + OpenMontage + GLM-5.2 31.3 Jun 2026 Official leaderboard
13 DeepSeek Harness + GLM-5.2 31.1 Jun 2026 Official leaderboard
14 OpenSquilla + Claude Opus 5 30.0 Jul 2026 Official leaderboard
15 deepseek-v4-vis 29.4 Claude Code + OpenMontage Official leaderboard
16 OpenSquilla + DeepSeek-V4-Flash 29.3 Apr 2026 Official leaderboard
17 Codex + OpenMontage + Claude Opus 5 28.9 Jul 2026 Official leaderboard
18 DeepSeek Harness + DeepSeek-V4-Flash 26.7 Apr 2026 Official leaderboard
19 OpenClaw + Claude Opus 5 25.8 Jul 2026 Official leaderboard
20 VideoClaw + Claude Opus 5 25.8 Jul 2026 Official leaderboard
21 Codex + OpenMontage + GLM-5.2 23.6 Jun 2026 Official leaderboard
22 OpenSquilla + GLM-5.2 21.6 Jun 2026 Official leaderboard
23 VideoClaw + GLM-5.2 18.0 Jun 2026 Official leaderboard
24 Codex + Claude Opus 5 17.3 Jul 2026 Official leaderboard
25 ViMax + Claude Opus 5 17.1 Jul 2026 Official leaderboard
26 OpenClaw + GLM-5.2 14.3 Jun 2026 Official leaderboard
27 UniVA + Claude Opus 5 11.1 Jul 2026 Official leaderboard
28 ViMax + GLM-5.2 8.80 Jun 2026 Official leaderboard
29 Codex + GLM-5.2 8.20 Jun 2026 Official leaderboard
30 OpenClaw + DeepSeek-V4-Flash 7.44 Apr 2026 Official leaderboard
31 Codex + DeepSeek-V4-Flash 6.48 Apr 2026 Official leaderboard
32 UniVA + GLM-5.2 5.62 Jun 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Claude Opus 5 51.8 Claude Code Jul 2026 Official leaderboard

Scores over time

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

In the benchmark’s paper
51.8 51.8

Source: Official VAHBench 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

All 695 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 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 3 pass 2 notes
  • Pass Independent runs Scores from the benchmark’s paper (32 entries).
  • Pass Models scored 1 model we track, plus 31 other entries on its leaderboard.
  • Pass Newest score The newest is from July 2026, for Claude Opus 5.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Claude Opus 5 (#11).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 30 rows for models we track and 2 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