Benchmark Almanac

Qwen2-VL-7B-Instruct

Qwen2-VL-7B-Instruct ranks first on VCR (Visual Caption Restoration).

Released 29 Aug 2024
28
Benchmarks with a score
1
Ranked first, out of five or more models
2
In the top three
Unknown lab Aug 2024

Qwen2-VL-7B-Instruct

No Epoch Capabilities Index score yet
  • OCRBench v2 42.3% #37 /65
  • VisualPuzzles 30.2% #21 /29
  • HumanEval-V 1.60% #18 /24

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 80.3% #43 of 178 Official leaderboard
Vectara Hallucination Leaderboard (HHEM) 4.20% #79 of 148 Official leaderboard
Creation-MMBench -43.76 #18 of 19 Official leaderboard

Math and science

BenchmarkScoreRankSource
U-MATH 26.3% #28 of 30 Official leaderboard

Coding

BenchmarkScoreRankSource
HumanEval-V 1.60% #18 of 24 Official leaderboard

Agents and tools

BenchmarkScoreRankSource
ScreenSpot-Pro 1.58% #10 of 11 Official leaderboard
BALROG 3.70% #35 of 35 Official leaderboard

Vision, audio and robotics

BenchmarkScoreRankSource
VCR (Visual Caption Restoration) 75.6% #1 of 42 Official leaderboard
A-Bench 76.4% #19 of 116 Official leaderboard
CCBench (MMBench Chinese Culture) 65.7% #30 of 152 Official leaderboard
MMBench 82.8% #33 of 155 Official leaderboard
MMBench-EN v1.1 81.8% #40 of 178 Official leaderboard
TOMATO 31.5% #5 of 22 Official leaderboard
MTVQA-Test 25.6% #5 of 18 Official leaderboard
OVBench 49.7% #2 of 7 Official leaderboard
MMStar 60.7% #57 of 173 Official leaderboard
MEGA-Bench 36% #9 of 22 Official leaderboard
CharXiv 34.6% #33 of 77 Official leaderboard
VHELM 53.4 #19 of 41 Official leaderboard
Video-MME 63.9% #24 of 50 Epoch AI
OCRBench v2 42.3% #37 of 65 Official leaderboard
MME-CoT 42.1% #7 of 11 Official leaderboard
MMVU 42.1% #18 of 28 Official leaderboard
ERQA 34.4% #12 of 17 Official leaderboard
VisualPuzzles 30.2% #21 of 29 Official leaderboard
Vision Arena 1,031 #123 of 136 Official leaderboard
MATH-Vision (MathVision) 16.3% #95 of 105 Official leaderboard

Other

BenchmarkScoreRankSource
OpenVLM Leaderboard 67.1 #49 of 173 Official leaderboard