Weekly coding model ranking, by BenchLM

This week's top 10 models on BenchLM's coding ranking, renewed every week.

RankingWeeklyAs of 2026-10-29

Korean version of this post View as Markdown

1Claude Opus 5.5Anthropic82.9Official site
2Claude Sonnet 5.5Anthropic81.7Official site
3Gemini 4 CarbonGoogle · Not yet listed in Google's model docs81.1Official site

4Claude Fable 5.1Anthropic76.4Official site
5GPT-6 AstraOpenAI74.7Official site
6Gemini 4 ArgonGoogle72.2Official site
7Claude Fable 5Anthropic71.4Official site
8Muse Wonder FullMeta71.1Official site
9Claude Opus 5Anthropic70.5Official site
10GPT-6.1 SolOpenAI70.3Official site

Scores are copied as the source published them. The same model name can score differently depending on the agent tool it runs in, so rankings differ from source to source. Each button goes to the model's official page.

AutomationTrying AI Design Skills, a rule pack for coding agentsThe GitHub repository AI Design Skills, with 10,166 stars, gives AI coding tools a set of landing page design rules. Here is how to install it, how it is used, and two samples built from two prompts.
AutomationWho is OpenAI GPT-6.1 Terra forOpenAI released GPT-6.1 Terra, a mid-range model. Here is where a team that automates repetitive work could use it.
AutomationFour things to check before you upgrade to n8n v3.0n8n, the workflow automation tool, shipped v3.0 as a pre-release. Here are four things from the release notes to check before you move.

Keep reading