VLM-R1-3B-Math-0305 — benchmark results

Provider: Other. Released 2025-03-05. Access: Open.

Unified ELO 1486 ± 10, rank #898 of 1841 rated models, from 98 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open LMM Reasoning - LogicVista - spatial33.3Accuracy (%)93.8
OpenVLM MathVista - LOG35.1Accuracy (%)90.3
Open LMM Reasoning - WeMath - Correspondence of Coordinates and Positions83.3Accuracy (%)90.1
Open LMM Reasoning - MathVerse - Analytic46.5Accuracy (%)85.7
OpenVLM MathVista - FQA72.1Accuracy (%)85.4
OpenVLM MathVista - VQA57.5Accuracy (%)84.7
Open LMM Reasoning - MathVision - descriptive geometry25Accuracy (%)84.2
OpenVLM MathVista - STA79.1Accuracy (%)78.7
Open LMM Reasoning - WeMath - Angles and Length48.4Accuracy (%)75.3
OpenVLM MathVista - SCI65.6Accuracy (%)71.6
Open LMM Reasoning - MathVision - transformation geometry26.2Accuracy (%)69.4
Open LMM Reasoning - WeMath - InadequateGeneralization (Loose)13.1Accuracy (%)67.3

Interactive version: theaggregate.ai/model?slug=vlm-r1-3b-math-0305 · How It Works · Data refreshed daily, snapshot 2026-07-25.