Benchmark Almanac

Qwen2-VL-2B-Instruct

Qwen2-VL-2B-Instruct places in the top three on VCR (Visual Caption Restoration).

Qwen Released August 29th, 2024 Family Qwen2-VL The lab's model card Compare with others
22
Benchmarks with a score
0
Ranked first, out of five or more models
1
In the top three
Qwen Aug 2024

Qwen2-VL-2B-Instruct

Open weights Small

No Epoch Capabilities Index score yet
  • Video SimpleQA 17.2% #21 /27
  • HR-Bench 61.3% #6 /8
  • VisualPuzzles 31.3% #18 /29

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.

Knowledge and language

BenchmarkScoreRankSource
MMBench-CN v1.1 71.2% #95 of 178 Official leaderboard
Video SimpleQA 17.2% #21 of 27 Official leaderboard
Vectara Hallucination Leaderboard (HHEM) 8.30% #122 of 148 Official leaderboard

Coding

BenchmarkScoreRankSource
ChartMimic 22.8 #10 of 14 Official leaderboard

Vision, audio and robotics

BenchmarkScoreRankSource
VCR (Visual Caption Restoration) 61.3% #2 of 42 Official leaderboard
CCBench (MMBench Chinese Culture) 53.7% #55 of 152 Official leaderboard
MM-Vet 51.5% #84 of 164 Official leaderboard
A-Bench 71.3% #61 of 116 Official leaderboard
MMBench 74.6% #83 of 155 Official leaderboard
SEED-Bench-2-Plus 61.2% #73 of 136 Official leaderboard
MMBench-EN v1.1 73.2% #96 of 178 Official leaderboard
TOMATO 26% #13 of 22 Official leaderboard
VisualPuzzles 31.3% #18 of 29 Official leaderboard
MMStar 47.5% #114 of 173 Official leaderboard
CoreCognition 51.7% #83 of 119 Official leaderboard
HR-Bench 61.3% #6 of 8 Official leaderboard
CharXiv 24.5% #58 of 77 Official leaderboard
CharXiv (DQ) 44.5% #53 of 67 Official leaderboard
MEGA-Bench 22.3% #18 of 22 Official leaderboard
MMVU 36.5% #23 of 28 Official leaderboard
MATH-Vision (MathVision) 12.4% #100 of 105 Official leaderboard

Other

BenchmarkScoreRankSource
OpenVLM Leaderboard 57.3 #98 of 173 Official leaderboard