BenchLeader

Most used in coding agents

The models people ran most inside coding agents (Claude Code, Codex, OpenCode and others) over the last 30 days, by share of turns in T3 Code's usage telemetry. One client's users, not the whole market.

As of 9 Nov 2026, Claude Opus 5.5 by Anthropic leads this list (57.1%), ahead of GPT-6.1 Sol (19.2%).

26 models qualify · data as of 9 Nov 2026

26 of 26
#
1Claude Opus 5.5max57.1%68.9$8.00115
2GPT-6.1 Solmax19.2%67.0$4.0057
3GPT-6 Astra5.1%68.1$20.0027
4Claude Sonnet 5.5max3.8%65.0$4.00152
5Claude Fable 5.1max2.0%66.9$20.0087
6GPT-6 Lunamax1.3%56.5$0.200104
7GPT-6.1 Terramax0.9%65.0$2.0073
8Claude Haiku 5.5max0.9%56.8$0.200238
9GPT-6 Solmax0.2%62.9$4.00121
10GPT-5.6 Solmax0.2%65.4$8.0099
11Claude Opus 50.1%65.2$10.0064
12GPT-5.6 Lunaxhigh0.1%58.4$0.450102
13Gemini 3.8 Flashhigh0.1%61.8$1.50112
14Claude Opus 4.8high0.1%60.5$10.0071
15GPT-5.6 Terramax0.1%60.5$4.50106
16Claude Sonnet 5high0.1%59.2$4.00105
17Claude Sonnet 4.60.0%58.2$6.0044
18Claude Opus 4.6high0.0%62.7$10.0039
19Claude Fable 50.0%66.6$20.0055
20Claude Haiku 4.50.0%47.6$2.0090
21GPT-5.5xhigh0.0%64.3$11.2580
22GPT-5.4xhigh0.0%63.2$5.6373
23Claude Opus 4.70.0%63.2$10.0048
24Claude Opus 4.5thinking0.0%59.7$10.0067
25GPT-5.4 Minixhigh0.0%56.0$1.69198
26GPT-5.1-Codexhigh0.0%–$3.4498