MMODELYST
CAP

Code

League table — best at code · 30
mean percentile across this capability's benchmarks
#ModelLeagueCoverage$/1Mtok/s
1Claude Fable 5100.01/4$2065
2Gemini 3.1 Pro Preview99.81/4$4.5140
3Gemini 3 Pro Preview99.52/4$4.5
4Muse Spark 1.199.51/4$2129
5GPT-5.499.31/4$5.63161
6GPT-5.6 Sol (max)98.81/4$11.2585
7GPT-5.598.51/4$11.2572
8GPT-5.6 Sol98.31/4$11.2571
9GPT-5.2 Codex98.01/4$4.81144
10Claude Opus 4.797.81/4$1052
11Grok 4.597.51/4$3110
12GPT-5.6 Terra (max)97.31/4$5.63166
13Gemini 3 Flash Preview97.12/4$1.13202
14GPT-5.297.02/4$4.8181
15Claude Sonnet 597.01/4$479
16Kimi K2.696.81/4$1.7143
17Claude Opus 4.896.51/4$1059
18GPT-5.3 Codex96.31/4$4.8199
19Gemini 3.5 Flash96.01/4$3.38238
20GPT-5.6 Luna (max)95.81/4$2.25234
21Claude Opus 4.695.31/4$1048
22Claude Opus 4.595.22/4$1057
23GPT-5.6 Terra95.01/4$5.63130
24Muse Spark94.81/4
25GLM-5.2 (max)94.31/4$2.15200
26GPT-5.5 Instant (May 2026)94.01/4$11.25
27MiMo-V2.5-Pro93.81/4$0.54455
28GLM-4.793.72/4$1137
29Grok Build 0.1 061693.51/4$1.2559
30DeepSeek V4 Pro93.31/4$0.54464

League = a model's mean percentile across the 4 benchmarksin this capability (best score per benchmark; partial coverage shown). Price & speed via Artificial Analysis.

Benchmarks in this capability (4)
HumanEval16 scores
LiveCodeBench279 scores
MBPP0 scores
SciCode402 scores