Replit’s Model Selector now routes tasks across Kimi K3, DeepSeek V4 Flash and GPTs
amasad · x · 2026-07-29
Replit’s new Model Selector is positioned as a task-based router for AI work: choose the best model for the job while controlling intelligence, cost, and speed.
The post highlights a mix of model options, including Kimi K3, DeepSeek V4 Flash, OpenAI GPT models, Anthropic models, and open-weight models, framing the product as a way to optimize outcomes rather than lock users into one provider.
More from coding & agent
- tldraw ships an offline app with Agent Skills for Claude and Codex drawing — vista8 · 2026-07-29
- A Claude agent scaffold organizes rules, skills, memory, and journaling — RileyRalmuto · 2026-07-29
- Jack Dorsey open-sources a self-hosted AI agent framework with 14.4K GitHub stars — goyalshaliniuk · 2026-07-29
- Google Gemini CLI nightly adds Firestore dual-locking and a new agent runner — gemini-cli-robot · 2026-07-29
- How to use GPT with Blender for planning and generating a school scene — remybigot · 2026-07-29
- How do you evaluate whether an agent change really improved anything? — Substantial_Step_351 · 2026-07-29