Benchmark Almanac

MediaPerf

GPT-5.4 0.42 leads since March 2026
Not in model cards

Models tag and summarize video advertisements. The benchmark measures tag accuracy, summary quality, cost and throughput.

The September release uses 2,003 videos. It scores tagging with macro F1 and grades summaries with a model.

Made by
Coactive AI
Released
March 2026
Kind
Video understanding
Level
University
Scored
Checked automatically
Answers
Open-ended
Tasks
All public
Measures
Capability
Language
English
In and out
Video → Text
Interaction
Single turn
Items from
Real work and use
Run by
Anyone
In views
Vision
Strongest model tested
GPT-5.4 #28 on the capability index
Last model added
April 2026
Scores refreshed
November 20th, 2026
17
Entries on its own leaderboard, 16 models we track

GPT-5.4 leads with 0.42 among models released by April 2026 GPT-5.4 leads with 0.42 among models released by April 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The default Video Tagging table ranks models by macro F1 on the 68 core tags. 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:

  1. 1 GPT-5.4 0.42
  2. 2 Gemini 3.1 Flash-Lite 0.41
  3. 3 Claude Opus 4.6 0.41
  4. 4 GPT-5.1 0.39
  5. 5 Gemini 3.1 Pro 0.39
  6. 6 Claude Sonnet 4.6 0.38
  7. 7 Nova 2 Lite 0.38
  8. 8 GPT-5 mini 0.37
  9. 9 Claude Haiku 4.5 0.37
  10. 10 Gemini 2.5 Pro Jun 2025 0.34
  11. 11 Gemini 3 Pro 0.34
  12. 12 Nemotron 3 Nano Omni 30B-A3B FP8 0.31
  13. 13 Qwen3-VL-30B-A3B-Instruct FP8 0.31
  14. 14 Nemotron Nano 12B v2 VL 0.30
  15. 15 Pegasus 1.2 0.30
  16. 16 GPT-5 nano 0.27
  17. 17 Amazon Nova Pro 0.25

Source: Official MediaPerf leaderboard

#EntryScoreSettingDateSource
1 GPT-5.4 0.42 Mar 2026 Official leaderboard
2 Gemini 3.1 Flash-Lite 0.41 Mar 2026 Official leaderboard
3 Claude Opus 4.6 0.41 Feb 2026 Official leaderboard
4 GPT-5.1 0.39 Nov 2025 Official leaderboard
5 Gemini 3.1 Pro 0.39 Feb 2026 Official leaderboard
6 Claude Sonnet 4.6 0.38 Feb 2026 Official leaderboard
7 Nova 2 Lite 0.38 Dec 2025 Official leaderboard
8 GPT-5 mini 0.37 Aug 2025 Official leaderboard
9 Claude Haiku 4.5 0.37 Oct 2025 Official leaderboard
10 Gemini 2.5 Pro Jun 2025 0.34 Jun 2025 Official leaderboard
11 Gemini 3 Pro 0.34 Nov 2025 Official leaderboard
12 Nemotron 3 Nano Omni 30B-A3B 0.31 FP8 Apr 2026 Official leaderboard
13 Qwen3-VL-30B-A3B-Instruct 0.31 FP8 Oct 2025 Official leaderboard
14 Nemotron Nano 12B v2 VL 0.30 Oct 2025 Official leaderboard
15 Pegasus 1.2 closed weights 0.30 Official leaderboard
16 GPT-5 nano 0.27 Aug 2025 Official leaderboard
17 Amazon Nova Pro 0.25 Dec 2024 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-5.4 0.42 Mar 2026 Official leaderboard
2 Gemini 3.1 Flash-Lite 0.41 Mar 2026 Official leaderboard
3 Claude Opus 4.6 0.41 Feb 2026 Official leaderboard
4 GPT-5.1 0.39 Nov 2025 Official leaderboard
5 Gemini 3.1 Pro 0.39 Feb 2026 Official leaderboard
6 Claude Sonnet 4.6 0.38 Feb 2026 Official leaderboard
7 Nova 2 Lite 0.38 Dec 2025 Official leaderboard
8 GPT-5 mini 0.37 Aug 2025 Official leaderboard
9 Claude Haiku 4.5 0.37 Oct 2025 Official leaderboard
10 Gemini 2.5 Pro Jun 2025 0.34 Jun 2025 Official leaderboard
11 Gemini 3 Pro 0.34 Nov 2025 Official leaderboard
12 Nemotron 3 Nano Omni 30B-A3B 0.31 FP8 Apr 2026 Official leaderboard
13 Qwen3-VL-30B-A3B-Instruct 0.31 FP8 Oct 2025 Official leaderboard
14 Nemotron Nano 12B v2 VL 0.30 Oct 2025 Official leaderboard
15 GPT-5 nano 0.27 Aug 2025 Official leaderboard
16 Amazon Nova Pro 0.25 Dec 2024 Official leaderboard

Rankings by part

We read 3 parts from the benchmark’s own leaderboard: Video Tagging: Core dataset, Video Tagging: Extended dataset and Video Summarization: Core dataset. Each ranks models on one part of the test. GPT-5.4 leads 1 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Video Tagging: Core dataset

16 models we track have a score here.

  1. 1 GPT-5.4 0.42
  2. 2 Gemini 3.1 Flash-Lite 0.41
  3. 3 Claude Opus 4.6 0.41
  4. 4 GPT-5.1 0.39
  5. 5 Gemini 3.1 Pro 0.39
  6. 6 Claude Sonnet 4.6 0.38
  7. 7 Nova 2 Lite 0.38
  8. 8 GPT-5 mini 0.37
  9. 9 Claude Haiku 4.5 0.37
  10. 10 Gemini 2.5 Pro Jun 2025 0.34

Quality vs. Cost

The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

0.3 0.4 0.5 $0$2.5$5$7.5$10$12.5 Cost per hour of video ($), lower is better F1 (macro) GPT-5.4Gemini 3.1 Flash-Lite 0.3 0.4 0.5 $0$2.5$5$7.5$10$12.5 Cost per hour of video ($), lower is better F1 (macro) GPT-5.4Gemini 3.1 Flash-Lite
ModelSettingF1 (macro)Cost per hour of video
GPT-5.4 0.42 $7.34
Gemini 3.1 Flash-Lite 0.41 $0.14
Claude Opus 4.6 0.41 $10.92
GPT-5.1 0.39 $2.04
Gemini 3.1 Pro 0.39 $2.11
Claude Sonnet 4.6 0.38 $6.57
Nova 2 Lite 0.38 $0.54
GPT-5 mini 0.37 $0.64
Claude Haiku 4.5 0.37 $2.42
Gemini 2.5 Pro Jun 2025 0.35 $2.13
Gemini 3 Pro 0.34 $3
Nemotron 3 Nano Omni 30B-A3B FP8 0.31 $0.88
Qwen3-VL-30B-A3B-Instruct FP8 0.31 $0.68
Nemotron Nano 12B v2 VL 0.3 $1.11
Pegasus 1.2 0.3 $1.85
GPT-5 nano 0.27 $0.18
Amazon Nova Pro 0.25 $1.37

Quality vs. Cost

The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

0.3 0.4 0.5 $0$2.5$5$7.5$10$12.5 Cost per hour of video ($), lower is better F1 (macro) GPT-5.4Gemini 3.1 Flash-Lite 0.3 0.4 0.5 $0$2.5$5$7.5$10$12.5 Cost per hour of video ($), lower is better F1 (macro) GPT-5.4Gemini 3.1 Flash-Lite
ModelSettingF1 (macro)Cost per hour of video
GPT-5.4 0.42 $7.34
Gemini 3.1 Flash-Lite 0.41 $0.14
Claude Opus 4.6 0.41 $10.92
GPT-5.1 0.39 $2.04
Gemini 3.1 Pro 0.39 $2.11
Claude Sonnet 4.6 0.38 $6.57
Nova 2 Lite 0.38 $0.54
GPT-5 mini 0.37 $0.64
Claude Haiku 4.5 0.37 $2.42
Gemini 2.5 Pro Jun 2025 0.35 $2.13
Gemini 3 Pro 0.34 $3
Nemotron 3 Nano Omni 30B-A3B FP8 0.31 $0.88
Qwen3-VL-30B-A3B-Instruct FP8 0.31 $0.68
Nemotron Nano 12B v2 VL 0.3 $1.11
GPT-5 nano 0.27 $0.18
Amazon Nova Pro 0.25 $1.37

3 models have held number one on the official leaderboard since its release in March 2026

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

Anthropic Google DeepMind OpenAI
Own leaderboard Not tested since Apr 2026
GPT-5.4 0.42

Source: Official MediaPerf leaderboard

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
On the benchmark’s own leaderboard
0.25 0.30 0.35 0.40 0.45 2025 2026 0.25 0.30 0.35 0.40 0.45 2025 2026

Source: Official MediaPerf 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 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 (17 entries).
  • Pass Models scored 16 models we track, plus 1 other entry 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-5.4 (#28).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on mediaperf.org: 16 rows for models we track and 1 other entry 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