# Per-model reasoning effort, GPT-5.5, and Gemini 3.5 Flash

> The model picker shows an estimated price and an effort control using each provider's native levels. GPT-5.5, Gemini 3.5 Flash, and Kimi K2.6 arrived.

May 28, 2026. Topics: Models. https://www.generalinput.com/changelog/2026-05-28-model-effort-and-new-models

Choosing a model is two decisions: which model, and how hard it should think. The model picker now shows both, along with what each option costs, and five new models arrived during May. Compare them on the [models page](/models).

### How it works

- **Estimated price.** Models are grouped into Basic, Capable, and Advanced tiers, each showing an estimated price in credits per 1,000 tokens, weighted 10 to 1 toward input because agent turns read far more than they write.
- **Native effort levels.** The effort picker lists only the levels the model's provider accepts (low to xhigh for GPT, low to max for Claude Sonnet and Opus, minimal to high for Gemini Flash) and passes the value straight through.
- **Saved defaults.** Your default model is saved with its effort, separately for chat and for building workflows.
- **Switch mid-conversation.** Conversations can move between providers, with tool-call history normalized so the next provider accepts it.

### New models

GPT-5.5, Gemini 3.5 Flash, Kimi K2.6, Qwen 3.6 Plus, and GLM 5.1. Most open-weight models moved to Fireworks AI, and saved selections moved with them, so no workflow had to be reconfigured.