Skip to content

Benchmark

FrontierCode 1.1

Coding benchmarks compare models on programming tasks. Scores are percentages.

Models measured
41
Scores recorded
102
Sources
2
IQ weight
2.97

Score distribution

45.2%
Mean
47.5%
Median
12.5%
Std dev
27%
10th percentile
60.3%
90th percentile
9.3%
Min
9.3%65.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.97 for chat models, measured over 41 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model
51Claude Haiku 5.5main· FrontierCode data46.4%
52GLM 5.3 Flashextended· FrontierCode data46.2%
53kimi-k2-7extended· FrontierCode data45.4%
54DeepSeek V4 Pro 0813extended· FrontierCode data44.5%
55Kimi K3main· FrontierCode data44.2%
56DeepSeek V4.1 Flashmain· FrontierCode data43.9%
57Claude Opus 4.6extended· FrontierCode data43.7%
58Gemini 3.7 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim43.6%
59Gemini 3.7 Flashmain· FrontierCode data43.6%
60GPT-5.4 miniextended· FrontierCode data43%
61GPT 5.5· LLM Stats data43%
62GPT 5.5main· FrontierCode data43%
63Claude Sonnet 5main· FrontierCode data42.7%
64Claude Sonnet 5· LLM Stats data42.7%
65Grok 4.5main· FrontierCode data42.4%
66GPT-6 Lunamain· FrontierCode data42.4%
67GPT-6 Luna· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim42.4%
68Grok 4.5· LLM Stats data42.4%
69GPT-5.6 Terramain· FrontierCode data41.3%
70GPT-5.6 Terra· LLM Stats data41.3%
71Gemini 3.8 Flashmain· FrontierCode data41.2%
72GLM 5.3main· FrontierCode data40.1%
73GLM 5.2extended· FrontierCode data40%
74Claude Sonnet 4.6extended· FrontierCode data40%
75GPT-5.6 Lunamain· FrontierCode data39.8%
76GPT-5.6 Luna· LLM Stats data39.8%
77Claude Opus 4.7main· FrontierCode data38.5%
78Claude Opus 4.7· LLM Stats data38.5%
79Gemini 3.6 Flashmain· FrontierCode data34.4%
80GLM 5.3 Flashmain· FrontierCode data31.8%
81DeepSeek V4 Flash 0731extended· FrontierCode data31.7%
82DeepSeek V4 Proextended· FrontierCode data31%
83Kimi K2.7 Code· LLM Stats data30.1%
84kimi-k2-7main· FrontierCode data30.1%
85DeepSeek V4 Pro 0813main· FrontierCode data28.6%
86MiniMax M3extended· FrontierCode data28.5%
87Nemotron 3 Ultraextended· FrontierCode data27.2%
88GPT-5.4 minimain· FrontierCode data27%
89Claude Opus 4.6main· FrontierCode data26.6%
90Inkling FP4extended· FrontierCode data24.8%
91GLM 5.2· LLM Stats data24.5%
92GLM 5.2main· FrontierCode data24.5%
93Claude Sonnet 4.6main· FrontierCode data24.3%
94Qwen3.7 Plusextended· FrontierCode data21.8%
95DeepSeek V4 Flash 0731main· FrontierCode data18.8%
96DeepSeek V4 Promain· FrontierCode data17.6%
97MiniMax M3· LLM Stats data14.7%
98MiniMax M3main· FrontierCode data14.7%
99Inkling FP4main· FrontierCode data14%
100Nemotron 3 Ultramain· FrontierCode data13.6%

Showing 51–100 of 102 results.

Tokens the model generated per task, as measured by the source. Comparable within this benchmark only.

Data credits: Data by LLM Stats