Windsurf Ships Adaptive Model Picker, GPT-5.4 Support, and Token Transparency in Response Cards
Releasebot·medium signal
Windsurf's new Adaptive model dynamically selects the best model per task to stretch quota without manual switching. GPT-5.4 arrives with limited-time self-serve pricing across reasoning tiers. Response cards now show token rates, prompt cache timing, and per-response token counts — a transparency play that no other AI IDE currently offers at this granularity.