Skip to content

Benchmark

SVGBench

Coding benchmarks compare models on programming tasks. Scores are percentages. Higher is better.

Models measured
82
Scores recorded
88
Sources
3
IQ weight
2.21

Score distribution

47.1%
Mean
48.2%
Median
14.6%
Std dev
28.4%
10th percentile
64.1%
90th percentile
6.8%
Min
6.8%74.4%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.21 for chat models, measured over 82 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51Mistral Medium 3.1svg-generation· SVGBench data46.1%
52Mistral Large 3svg-generation· SVGBench data46%
53Grok 3svg-generation· SVGBench data45.8%
54GPT OSS 120Bsvg-generation· SVGBench data45.6%
55Gemini 2.5 Flash-Litesvg-generation· SVGBench data43.7%
56Kimi K2svg-generation· SVGBench data43.6%
57o3 Minisvg-generation· SVGBench data42.7%
58Gemini 2.5 Flash Lite Preview 09-2025svg-generation· SVGBench data41.8%
59Gemini 2.0 Flashsvg-generation· SVGBench data41.6%
60Qwen3 VL 235Bsvg-generation· SVGBench data41%
61Qwen3 30B A3B Instructsvg-generation· SVGBench data40.9%
62DeepSeek R1svg-generation· SVGBench data40.4%
63Gemini 2.0 Flash Litesvg-generation· SVGBench data39.6%
64Qwen3 Next 80B A3Bsvg-generation, thinking· SVGBench data39.2%
65Qwen3 Coder 30B A3B Instructsvg-generation· SVGBench data38%
66Codestral 25.01svg-generation, 2508· SVGBench data37.9%
67Llama 4 Mavericksvg-generation· SVGBench data36%
68Mistral Small 3.2 24B Instructsvg-generation· SVGBench data35.9%
69GPT-4.1 nanosvg-generation· SVGBench data35.5%
70GPT OSS 20bsvg-generation· SVGBench data34.7%
71Nova 2 Litesvg-generation, non-thinking· SVGBench data33.9%
72Qwen 3 235B A22Bsvg-generation· SVGBench data33.9%
73Claude Haiku 3.5svg-generation· SVGBench data33.3%
74Nova 2 Litesvg-generation, thinking· SVGBench data32%
75GPT 4osvg-generation· SVGBench data31.8%
76Hermes 4 405Bsvg-generation· SVGBench data31.6%
77Gemma 3 27B ITsvg-generation· SVGBench data29.6%
78Qwen3 32Bsvg-generation· SVGBench data29.4%
79Qwen3 30B A3Bsvg-generation, 2507, thinking· SVGBench data28.4%
80Llama 4 Scoutsvg-generation· SVGBench data26.9%
81Qwen3 30B A3Bsvg-generation· SVGBench data26.1%
82Gemma 3 12B ITsvg-generation· SVGBench data25.7%
83GPT-4o minisvg-generation· SVGBench data25.4%
84Hermes 4 70Bsvg-generation· SVGBench data21.6%
85Seed OSS 36B Instructsvg-generation· SVGBench data21.1%
86Gemma 3 4B ITsvg-generation· SVGBench data14.7%
87Trinity Minisvg-generation· SVGBench data8.7%
88Llama 3.1 8B Instructsvg-generation· SVGBench data6.8%

Showing 51–88 of 88 results.

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Data by LLM Stats