Skip to content

Benchmark

Kagi LLM Benchmark

Scores are percentages.

Models measured
110
Scores recorded
123
Sources
1
IQ weight
1.63

Score distribution

54.3%
Mean
55.2%
Median
17.8%
Std dev
31.5%
10th percentile
75%
90th percentile
6.3%
Min
6.3%91.4%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.63 for chat models, measured over 110 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model
51GPT OSS 120B· Kagi LLM Benchmark58.6%
52cogito-v2-preview-deepseek-671b· Kagi LLM Benchmark58.4%
53Claude Sonnet 4.5· Kagi LLM Benchmark57.9%
54GLM 4.5· Kagi LLM Benchmark57.9%
55Ling 1t· Kagi LLM Benchmark57.8%
56MiniMax M2· Kagi LLM Benchmark57.8%
57DeepSeek V3.1 Terminus· Kagi LLM Benchmark57.4%
58Gemini 2.5 Flash· Kagi LLM Benchmark56.8%
59Claude Sonnet 4· Kagi LLM Benchmark55.9%
60Llama 4 Maverick· Kagi LLM Benchmark55.9%
61Hermes 4 405Bthinking· Kagi LLM Benchmark55.8%
62MiniMax M2.5· Kagi LLM Benchmark55.2%
63Qwen 3 235B A22Bno-thinking· Kagi LLM Benchmark55%
64Qwen3 30B A3B2507, thinking· Kagi LLM Benchmark54.9%
65Qwen3 Next 80B A3B Instruct· Kagi LLM Benchmark54.4%
66DeepSeek V4 Pro· Kagi LLM Benchmark53.5%
67DeepSeek V3.1· Kagi LLM Benchmark53.2%
68GPT OSS 20b· Kagi LLM Benchmark53.2%
69Grok Code Fast 1· Kagi LLM Benchmark52.5%
70DeepSeek R1 Distill Llama 70b· Kagi LLM Benchmark52.3%
71DeepSeek V3· Kagi LLM Benchmark52.3%
72GPT 4.1· Kagi LLM Benchmark52.3%
73DeepSeek V3.2 Exp· Kagi LLM Benchmark52.2%
74DeepSeek V4 Flash· Kagi LLM Benchmark52.2%
75GLM 5· Kagi LLM Benchmark51.7%
76Nemotron-nano 12b v2-vl· Kagi LLM Benchmark51.7%
77GPT-5.6 Terra· Kagi LLM Benchmark51.3%
78Mistral Large 3· Kagi LLM Benchmark50.9%
79Mistral Medium· Kagi LLM Benchmark50%
80Qwen3 Coder· Kagi LLM Benchmark49.5%
81GPT-5.6 Luna· Kagi LLM Benchmark49.1%
82Qwen3 14B· Kagi LLM Benchmark49.1%
83Qwen3 32Bno-thinking· Kagi LLM Benchmark48.7%
84GPT-4.1 mini· Kagi LLM Benchmark48.6%
85Nemotron Nano 9b V2· Kagi LLM Benchmark47.4%
86GLM 4.6· Kagi LLM Benchmark45.7%
87ERNIE 4.5 300B A47B· Kagi LLM Benchmark45%
88Grok 4.20· Kagi LLM Benchmark44.8%
89Nova Premier 1.0· Kagi LLM Benchmark44.8%
90LongCat Flash Chat· Kagi LLM Benchmark43.9%
91GLM 4.5 Air· Kagi LLM Benchmark43%
92MiniMax 01· Kagi LLM Benchmark42.5%
93GLM 4 32B· Kagi LLM Benchmark42.3%
94Mistral Medium 3.5· Kagi LLM Benchmark41.4%
95Gemini 2.5 Flash-Lite· Kagi LLM Benchmark40.5%
96Mistral Small· Kagi LLM Benchmark40.5%
97GPT 5.4 Nano· Kagi LLM Benchmark39.7%
98GPT-5.4 mini· Kagi LLM Benchmark37.9%
99Phi 4 Reasoning Plus· Kagi LLM Benchmark37.8%
100Devstral Medium· Kagi LLM Benchmark37.7%

Showing 51–100 of 123 results.

Tokens per task, input and output together, as measured by the source. Comparable within this benchmark only.