diff --git a/README.md b/README.md index 57dd8b9aa..7d90c4b1f 100644 --- a/README.md +++ b/README.md @@ -38,6 +38,12 @@ bcode run "On Google flights return all flight details from New York to SF tomor ### Connect an LLM + + + + Best BrowserCode models on BU Bench + + @@ -48,11 +54,13 @@ BrowserCode supports any model you can reach with an API key, plus [every provid Use `/connect` in the TUI, or set provider API keys in your environment. -Recommended models from current BrowserCode evals: +Recommended models from current BU Bench evals: -- Frontier: `claude-opus-4-7`, `gpt-5.5` -- Value: `glm-5.1`, `mimo-v2.5-pro` -- Budget: `gemini-3-flash-preview` +- Best score: `claude-opus-4-7` +- Best speed: `gpt-5.5` +- Best open-weight: `glm-5.2` +- Best value: `gemini-3.1-pro-preview` +- Lowest cost: `gemini-3-flash-preview` ### Connect a Browser diff --git a/static/browsercode_best_models_dark.png b/static/browsercode_best_models_dark.png new file mode 100644 index 000000000..8351445bf Binary files /dev/null and b/static/browsercode_best_models_dark.png differ diff --git a/static/browsercode_best_models_light.png b/static/browsercode_best_models_light.png new file mode 100644 index 000000000..56ba1aef8 Binary files /dev/null and b/static/browsercode_best_models_light.png differ