ReportsAI Coding Tools
Industry report
AI Coding Tools
Which AI coding assistants and agents do AI models recommend? Recommendation rates from a fixed set of buyer prompts across ten models.
Last run Nov 15, 20262,500responses10 models
Rankings describe what AI models answered to a fixed set of prompts. They are not endorsements or quality ratings. Product names belong to their owners.
Model memory
Model memory — what each model generation learned in training. Not dated observations; no web search.
Coming soon — historical model backfill pending
Built to be repeated
Methodology v1- Fixed prompt set
- 50 prompts
- Topics
- 5 topics
- Sampling
- 5 samples / prompt / model
- Models
- 10 models, equal weight
- Minimum evidence
- 10 recommendations
Version 1. Editing a prompt starts a new series.
Equal weight. A topic with fewer answers does not count less.
2,500 responses in this run.
Claude Sonnet 5.5, DeepSeek V4.1 Flash, Gemini 4 Hydrogen, Muse Spark 1.3, Mistral Large 4, Kimi K3.1 Pro, GPT-6.1 Sol, Perplexity Sonar, Grok 4.7, GLM-5.3
Below that, the rank is insufficient evidence.
Recommendation rate is the share of valid responses that recommend a product, counted once per response. The all-models figure is the unweighted mean of the per-model rates, so a longer answer does not get a heavier vote. 95% intervals come from a bootstrap that resamples prompts, not individual answers. These answers come from provider APIs with pinned model slugs. They are not the consumer apps, which add their own system prompts, memory, and search. Lowest valid-response rate in this run: 100%.
“Which AI coding agents can take a GitHub issue and open a working pull request on their own?”
“What AI tools are best for debugging failing tests and fixing bugs in a Python project?”
Free audit
Where does your brand rank?
These reports cover a category. A free audit grades your own site against the same four models.