Skip to content

Benchmark

FrontierCode 1.0

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

Models measured
15
Scores recorded
30
Sources
1
IQ weight
1.76

Score distribution

28.2%
Mean
25.7%
Median
15.4%
Std dev
9.1%
10th percentile
46.8%
90th percentile
1.8%
Min
1.8%61.6%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.76 for chat models, measured over 15 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model
1Claude Fable 5extended· FrontierCode data61.6%
2Claude Opus 4.8extended· FrontierCode data51.8%
3Claude Fable 5main· FrontierCode data46.8%
4GPT 5.5extended· FrontierCode data44.8%
5Claude Opus 4.7extended· FrontierCode data43.9%
6kimi-k2-7extended· FrontierCode data39.9%
7GLM 5.2extended· FrontierCode data39%
8Kimi K2.6extended· FrontierCode data37%
9GPT-5.4 miniextended· FrontierCode data36%
10Claude Opus 4.8main· FrontierCode data34.3%
11Gemini 3.1 Pro Previewextended· FrontierCode data34.2%
12Claude Sonnet 4.6extended· FrontierCode data33.6%
13GPT 5.5main· FrontierCode data25.5%
14Claude Opus 4.7main· FrontierCode data23.8%
15Kimi K2.5extended· FrontierCode data22.7%
16kimi-k2-7main· FrontierCode data22%
17Qwen3.6 Plusextended· FrontierCode data21.3%
18MiniMax M2.7extended· FrontierCode data19.9%
19GLM 5.2main· FrontierCode data19.2%
20GPT-5.4 minimain· FrontierCode data17.8%
21Gemini 3.1 Pro Previewmain· FrontierCode data16.7%
22Kimi K2.6main· FrontierCode data16%
23MiniMax M2.5extended· FrontierCode data15.8%
24Claude Sonnet 4.6main· FrontierCode data15.1%
25Gemini 3.1 Flash Liteextended· FrontierCode data14.6%
26Qwen3.6 Plusmain· FrontierCode data9.7%
27Kimi K2.5main· FrontierCode data6.9%
28MiniMax M2.7main· FrontierCode data6%
29MiniMax M2.5main· FrontierCode data5.3%
30Gemini 3.1 Flash Litemain· FrontierCode data4.8%

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