LLM Reference
active

MathVista

Metric: Accuracy (higher is better)Introduced: 2023

Mathematical reasoning benchmark combining visual contexts (charts, geometry diagrams, scientific figures) from 28 existing datasets with 6,141 total problems. High benchmark score alone doesn't make a model the right pick — weigh it against pricing, API availability, and release date.

Models ranked

5

tracked on this benchmark

Score band

86.8 – 74.8

best → lowest tracked

Snapshot trend

-11.20

Apr 16 → Jun 7 · 2 models

Leaderboard

Tracked models ranked by Accuracy (higher is better).

Compare candidates
#Model variant and provenanceScore
1
o3
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Jun 3, 2026Confidence: Not recordedSource

Notes: DAT-5571 vision ranking audit: llm-stats reports o3 leads MathVista at 86.8%.

86.8
2
Qwen3.6-35B-A3B
Version: miniHarness: Not recordedEvaluator: Not recordedObserved: Apr 16, 2026Confidence: Not recordedSource
86.4
3
o4-mini
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Jun 3, 2026Confidence: Not recordedSource

Notes: DAT-5571 vision ranking audit: llm-stats reports o4-mini at 84.3% on MathVista.

84.3
4
MiniCPM-V 4.6
Version: Thinking variant (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
75.6
5
Claude Opus 4.6
Version: MathVista accuracy (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
74.8

How to read this benchmark

This benchmark scores models where higher is better. Scores are useful for directional filtering and shortlisting — not for universal quality ranking. Prefer benchmarks closest to your workload, then validate the linked model pages for pricing, context window, and provider availability.

Trust this score when

  • There is a fresh timestamped snapshot (or multiple snapshots) for this benchmark.
  • The model list covers the same version family you can actually deploy today.
  • Top candidates overlap with your required routing and feature requirements.

Be cautious when

  • There is only one benchmark snapshot or the dataset appears stale.
  • The benchmark metric direction is opposite of your decision objective.
  • The score difference between options is narrow and likely within implementation variance.

FAQ

What does the MathVista benchmark measure?

Mathematical reasoning benchmark combining visual contexts (charts, geometry diagrams, scientific figures) from 28 existing datasets with 6,141 total problems. On this page it lists 5 tracked model variants where higher is better.

Is a higher MathVista score always better?

For this benchmark, higher is better. A high score helps you shortlist, but confirm pricing, context window, and provider availability on each model page before committing — the top scorer is not always the right pick for your workload or budget.

How current is this MathVista data?

This benchmark was last reviewed on Apr 15, 2026. The tracked score average moved -11.20 points across the last 3 snapshots.

Last reviewed: Apr 15, 2026