whatmodel.sh / video / Veo 3.1

Veo 3.1

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Veo 3.1 is a closed-weights video model from Google, released October 2025. Licence: proprietary.

68.7±1.2#17 of 93 video modelsfrom 20 benchmark results across 17 sources

$ how it compares

The video models ranked either side of Veo 3.1, out of 139 with a score, then more from Google.

#modelscorewhereRAMreleased
17Grok Imagine Video xAI69±2api $0.050/sec2026-01
18Kling 3.0 Pro Kuaishou69±2api $0.11/sec2026-02
19Veo 3.1 Google69±1api $0.40/sec2025-10
20Sora 2 Pro OpenAI68±2comparison only2025-10
21Veo 3.1 Fast Google68±1api $0.10/sec2025-10
more from Google
1Gemini Omni 1.1 Flash Google81±2api $0.10/sec2026-08
3Gemini Omni Flash Google80±2api $0.10/sec2026-05
25Veo 3 Fast Google65±2comparison only2025-07

see all video models · find the ones that fit your machine

$ what it's good at

text to video▆▄▃▄▆▅█▃▂▃▂▄ #10 of 68 · 31% of its score · 3 tests

Beats Grok Imagine Video here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
LMArena video1364.2 ✎▂▂▃▆█▆·▆▅▁▂▇ beats 31/41avg 1147▇▇▇▇▇▇▇▇▇▇ 26%typical
Cosmos HUE human evaluation, text-to-video (Cosmos 3 report)91.3% ✎█▄▄··▄▄▄▄▄·▄ beats 9/9easy · avg 85%▇ 2.5%maxed out
Sci-VBench testmini (expert human rating)56.0% ✎▅·▃█▃·▅▃▅▃▃▃ beats 11/14hard · avg 44%▇ 2.3%blunt · at its level
image to video▁▂·▂▅█▅▄▅▆▃▄ #16 of 50 · 45% of its score · 9 tests

Beats HappyHorse 1.1 here, though they rank higher overall. Trails Kling 3.0, Veo 3.1 Fast here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Arena image-to-video1398.1 ✎▁▁▁▁▄▇▃▆▃▃▄█ beats 28/40avg 1189▇▇▇▇▇▇▇ 17%typical
Rapidata physics realism (I2V, with reference)1077.0 ✎▂··▂▂▆█▆▂▂▆▄ beats 8/19avg 741▇▇▇▇ 10%sharp
Target-Bench v2 (VGGT planning score, I2V)21.0% ✎▃·▃···▃▃█·▃▃ beats 5/8very hard · avg 9.6%▇▇▇ 6.6%typical
VGI-Bench visual reasoning (I2V)32.0% ✎▄▄···▄▄█·█·▄ beats 2/8hard · avg 20%▇ 3.3%at its level
VWG-Bench think-with-video reasoning (overall)71.0% ✎█··▄······▄█ beats 5/5medium · avg 50%▇ 2.5%at its level
Cosmos HUE human evaluation, image-to-video (Cosmos 3 report)89.7% ✎▄·█··▄··▄▄▄█ beats 8/8easy · avg 84%▇ 2.2%maxed out
What-If World causal-intervention video generation (APEO paired average, Gemini 3.1 Pro judge)50.9% ✎█·▄····▄···█ beats 4/5hard · avg 27%▇ 1.5%blunt · at its level
V-ReasonBench (average pass@5)24.3% ✎█·····▄···▄▄ beats 2/4hard · avg 17%▇ 1.0%at its level · noisy
1 more test
testresultvs the others in its rundifficultycountswhy
Rapidata camera movement (I2V)917.9 ✎▄···▄▄▄█··▄█ beats 2/8avg 794▇ 0.5%noisy
other tests8 not in a task · 24% of its score
testresultvs the others in its rundifficultycountswhy
Rapidata relative camera movement (I2V, alignment Elo)1114.3 ✎█···▄▄██·█▄█ beats 6/12avg 661▇▇ 6.1%sharp · noisy
MindWorldBench intention accuracy (I2V mental-state reasoning)59.5% ✎▅·▃▅·█·▃···▃ beats 9/9hard · avg 21%▇▇ 5.0%at its level
SWITCH interactive-interface I2V (VLM-judge overall)41.8% ✎███··█·····█ beats 3/4very hard · avg 28%▇ 3.7%at its level
Rapidata image-to-video realism (overall Elo)1266.6 ✎▂······▃▂▃▃█ beats 7/12avg 871▇ 3.3%sharp · noisy
VideoArgus-Bench T2V (rubric-grounded score)86.6% ✎█▄·▄█▄▄····▄ beats 7/8easy · avg 82%▇ 1.5%blunt
VideoArgus-Bench TI2V (rubric-grounded score)83.6% ✎█··▄▄▄▄····▄ beats 5/6easy · avg 77%▇ 1.5%blunt
PhyGround physics human evaluation (Overall, TI2V)65.6% ✎█▄▄··▄···▄·█ beats 6/7medium · avg 57%▇ 1.5%blunt · at its level
Datapoint Video Bench (overall Elo)963.8 ✎▃▃·▅·▃·▃··▃█ beats 1/9avg 889▇ 1.5%blunt

$ where to run it

Through an API

api · $0.40/sec · ai.google.dev ↗ · price sourceUSD per generated video second at 720p WITH audio on Gemini Developer API. Only successfully generated videos billed; other resolutions differ. Product: Gemini Developer API. API model: veo-3.1-generate-preview. Settings: resolution=720p, audio=yes, tier=paid. Price basis: per second of successfully generated output. Currency: USD.

$ how the score was worked out · LMArena leaderboard dataset 43%, Rapidata/world-model-physics… 10%, Target-Bench: Can Video World… 7% and 14 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
LMArena leaderboard dataset2▇▇▇▇▇▇▇▇▇▇▇▇▇ 43%CC BY 4.0
Rapidata/world-model-physics published dataset-card rankings1▇▇▇ 10%CC-BY-4.0
Target-Bench: Can Video World Models Achieve Mapless Path Planning with Semantic Targets?1▇▇ 7%CC-BY-4.0
Rapidata/relative-camera-movement-human-preference-900k published dataset-card rankings1▇▇ 6%CC-BY-4.0
MindWorldBench: Evaluating Mental-State-to-Behavior Reasoning in Image-to-Video Generation (arXiv:2609.39147v1)1▇ 5%CC-BY-4.0
Cosmos 3: Omnimodal World Models for Physical AI (arXiv:2606.02800v4)2▇ 5%CC-BY-4.0
SWITCH: Benchmarking Modeling and Handling of Tangible Interfaces in Long-horizon Embodied Scenarios (arXiv:2511.17649v5)1▇ 4%CC-BY-SA-4.0
VGI-Bench: Probing Visual Intelligence in Video Generation Models1▇ 3%CC-BY-4.0
Rapidata/image-to-video-realism published dataset-card rankings1▇ 3%CC-BY-4.0
VideoArgus: Agentic Rubric-Grounded Unified Evaluation for Video Generation and Editing (arXiv:2608.05485v1)2▇ 3%CC-BY-4.0
From Evaluation to Enhancement: Benchmarking and Improving Think-with-Video Reasoning for Video Generative Models (VWG-Bench)1▇ 3%CC-BY-4.0
Sci-VBench: Evaluating Knowledge- and Reasoning-Intensive Video Generation in Science Domains1▇ 2%CC-BY-4.0
What-If World: A Causal Benchmark for General World Models in Embodied Scenarios (arXiv:2605.27589v1), Table 21▇ 2%CC BY 4.0
PhyGround: Benchmarking Physical Reasoning in Generative World Models (arXiv:2605.10806v1)1▇ 2%CC-BY-4.0
Datapoint Video Bench leaderboard (datapointai/text-to-video-human-preferences-326k dataset card)1▇ 1%CC-BY-4.0
V-ReasonBench: Toward Unified Reasoning Benchmark Suite for Video Generation Models1▇ 1%CC-BY-4.0
Rapidata/camera-movement published dataset-card rankings1▇ 0%CC-BY-4.0
prior · pull toward the middle▇ 1%
weighted average= 68.7

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json