How many tracked prints does the MATH-Vision family have?

MATH-Vision leaderboard — the family board.

3 TRACKED VARIANTS · 1 FIRST-CLASS BOARD

One family, 3 tracked prints. Scoring configs and splits of the base board render as sections of this page, each under its own honest label.

● LIVE DATA53 MODELS ON THIS BOARDLEADER KIMI K3 AT 97.8

The board — MATH-VisionTOP 53 OF 53 MEASURED · BEST SCORE PER MODEL, DEFAULT VARIANT

#ModelVendorScoreVLast seen

01Kimi K3Moonshot97.82026-08-24

02GPT-5.6 SolOpenAI95.82026-08-24

03Qwen3.8 Flash NextAlibaba95.72026-08-27

04Claude Fable 5Anthropic94.82026-08-24

05Qwen3.8 27BAlibaba94.62026-08-26

06Seed 2.1 Pro PreviewByteDance94.52026-09-26

07Kimi K2.6Moonshot93.22026-08-23

08GPT-5.4OpenAI922026-05-03

09Qwen3.7 Plus PreviewAlibaba90.32026-08-26

10Gemini 3.1 ProGoogle89.82026-05-03

11Qwen3.6 PlusAlibaba882026-08-23

12Claude Opus 4.8Anthropic86.72026-08-24

13Qwen3.5 122B A10BAlibaba86.22026-08-23

14Gemini 3 ProGoogle86.12026-06-15

15Qwen3.5 27BAlibaba862026-08-23

16Gemma 4 31BGoogle85.62026-10-08

17Qwen3.6 27BAlibaba85.12026-08-24

18Kimi K2.5Moonshot84.22026-08-23

19Qwen3.5 35B A3BAlibaba83.92026-08-23

20GPT-5.2OpenAI832026-06-15

21Gemma 4 26B A4BGoogle82.42026-10-08

22Seed 1.8ByteDance81.32026-09-09

23Gemma 4 12BGoogle79.72026-10-08

24Claude Opus 4.5Anthropic77.12026-06-15

25Step3 VL 10BStepFun76.02026-08-23

26Qwen3 VL 235B A22B ReasoningAlibaba74.62026-08-23

27Gemini 2.5 ProGoogle73.32026-06-05

28Claude Opus 4.6Anthropic71.22026-08-26

29Qwen3 VL 32B ReasoningAlibaba70.22026-08-23

30Seed 1.5 VLByteDance68.72026-06-05

31Qwen3 VL 235B A22B InstructAlibaba66.52026-08-23

32Qwen3 VL 30B A3B ReasoningAlibaba65.72026-08-23

33GLM-4.6V (106B-A12B)Z.ai63.52026-06-05

34Qwen3 VL 32B InstructAlibaba63.42026-08-23

35Qwen3 VL Thinking (8B)Alibaba62.72026-08-23

36Qwen3 VL 30B A3B InstructAlibaba60.22026-08-23

37Qwen3 VL 4B (Reasoning)Alibaba602026-08-23

38MiMo VL RL 2508 (7B)Xiaomi59.62026-06-05

39Gemma 4 E4BGoogle59.52026-10-08

40Kimi VL A3B ThinkingMoonshot56.92026-06-05

41GLM-4.6V-Flash (9B)Z.ai54.02026-06-05

42Qwen3 VL 8B InstructAlibaba53.92026-08-23

43Gemma 4 E2BGoogle52.42026-10-08

44InternVL-3.5 (8B)OpenGVLab52.02026-06-05

45Qwen3 VL 4B InstructAlibaba51.62026-08-23

46Gemma 3 27BGoogle462026-10-08

47Qwen2.5 VL 32B InstructAlibaba402026-08-26

48Qwen2.5 VL 72BAlibaba39.32026-09-02

49Gemma 3 12BGoogle32.12026-06-05

50GPT-4oOpenAI30.42026-08-24

51Qwen2 VL 72B InstructAlibaba25.92026-08-25

52Qwen2.5 Omni 7BAlibaba252026-08-23

53Kimi VL A3B InstructMoonshot21.72026-06-12

◆ IV — INDEPENDENTLY VERIFIED · ◆ AA — AGGREGATOR · ◆ VA — VENDOR · ◇ SA — SOURCE-ATTRIBUTED

Variants of this boardSCORING CONFIGS & SPLITS — NOT CITED STANDALONE, SO THEY LIVE HERE

MathVision (full) (Pass@1)

VARIANT5 models · best score per model

#ModelVendorScoreVLast seen

01Qwen2.5 VL 72BAlibaba38.12026-06-05

02Kimi VL A3B ThinkingMoonshot36.82026-06-05

03Gemma 3 27BGoogle35.52026-06-05

04Gemma 3 12BGoogle32.12026-06-05

05GPT-4oOpenAI30.42026-06-05

mathvision_with_python

VARIANT5 models · best score per model

#ModelVendorScoreVLast seen

01GPT-5.4OpenAI96.12026-05-03

02Gemini 3.1 ProGoogle95.72026-05-03

03Kimi K2.6Moonshot93.22026-05-03

04Kimi K2.5Moonshot852026-05-03

05Claude Opus 4.6Anthropic84.62026-05-03

THE FAMILY BOARD · EVERY PRINT LABELED · EVERY SCORE SOURCE-ATTRIBUTED ON ITS OWN LIVE BOARDAll leaderboards The frontier board Models by vendor