Kimi-VL-A3B-Thinking-2506

Kimi-VL-A3B-Thinking-2506 places in the top three on MM-Vet.

Moonshot AI Released 21 Jun 2025 Family Kimi-VL The lab's model card

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Moonshot AI Jun 2025

Open weights Small

No Epoch Capabilities Index score yet

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Benchmark	Score	Rank	Source
MMBench-CN v1.1	79.8%	#46 of 178	Official leaderboard
MMLU	82%	#63 of 127	Model card
MMLU-Pro	68.5%	#117 of 151	Model card

Benchmark	Score		Rank	Source
MATH	91.8%		#7 of 93	Model card
GPQA Diamond	42.3%		#250 of 284	Model card

Benchmark	Score	Rank	Source
ScreenSpot-v2	91.4%	#9 of 14	Model card
ScreenSpot-Pro	52.8%	#21 of 25	Model card
OSWorld-G	52.5%	#5 of 5	Model card

Benchmark	Score	Rank	Source
MM-Vet	78.1%	#2 of 8	Model card
MMBench-EN v1.1	80.4%	#50 of 178	Official leaderboard
OCRBench	86.9%	#11 of 28	Model card
MathVista	80.1%	#26 of 64	Model card
MMMU	64%	#75 of 97	Model card
MATH-Vision (MathVision)	56.9%	#33 of 42	Model card
MMStar	70.4%	#24 of 30	Model card
RealWorldQA	70%	#25 of 30	Model card
V*	83.2%	#10 of 12	Model card
Video-MME	71.9%	#36 of 42	Model card
MMLongBench-Doc	42.1%	#12 of 13	Model card
MMMU-Pro	46.3%	#67 of 72	Model card
Video-MMMU	65.2%	#28 of 30	Model card
MMVU	57.5%	#20 of 20	Model card

Benchmark	Score		Rank	Source
OpenVLM Leaderboard	74.3		#15 of 173	Official leaderboard

The benchmarks the lab chose to headline, with the scores it gave.

Every result in the longer documents the lab published with it.