MathVision

MathVision - visual math problem solving benchmark. Reports Without CI and With CI (Consistency Inference) scores.

Category

Vision & Language

Max Score

100.0

Score Type

percent

Active

Yes

Model Scores

Scores for MathVision

Model Score Date Verified
Qwen3.8-27B
84.94%
18.08.2026 Verified
Qwen3.6-27B
75.48%
18.08.2026 Unverified
Qwen3.7-Plus
85.52%
18.08.2026 Unverified
Opus4.6 Max
37.64%
18.08.2026 Unverified
Qwen3.5-4B
55.21%
01.02.2026 Verified
Kimi K3
100.00%
18.08.2026 Verified
Qwen3.5-122B-A10B
77.61%
01.02.2026 Verified
Qwen3.5-27B
77.22%
01.02.2026 Unverified
Qwen3.5-35B-A3B
73.17%
01.02.2026 Unverified
Claude-Sonnet-4.5
48.46%
01.02.2026 Unverified
GPT-5-mini 2025-08-07
50.00%
01.02.2026 Unverified
Qwen3-235B-A22B
55.21%
01.02.2026 Unverified
Gemma4-31B-it
76.45%
02.07.2026 Verified
Gemma4-26B-A4B
70.27%
02.07.2026 Verified
Gemma4-12B
65.06%
02.07.2026 Verified
Gemma4-E4B
26.06%
02.07.2026 Verified
Gemma4-E2B
12.36%
02.07.2026 Verified
Gemma3-27B
46.00
02.07.2026 Unverified
Qwen3-VL-235B-A22B
55.21%
01.02.2026 Verified
Kimi K2.6
79.92%
23.08.2026 Verified
GPT-5.4
88.80%
23.08.2026 Verified
Claude Opus 4.6
48.65%
23.08.2026 Verified
Gemini 3.1 Pro
84.56%
23.08.2026 Verified
Kimi K2.5
73.75%
23.08.2026 Verified
Qwen3.8-Flash-Next
86.10%
26.08.2026 Verified
Claude-Opus-4.6 (Max)
37.64%
26.08.2026 Unverified