New conversations use Gemini 4 Hydrogen unless you have saved another model preference. Smart Auto remains an optional routing choice. The picker lists the following models; actual access requires the provider to be configured for the deployment. Selecting a model explicitly keeps that request on the selected provider; Kona reports unavailability instead of silently substituting another provider.
Smart Auto
Routes across current fast, lightweight, and frontier tiers by task complexity.
Gemini 4 Hydrogen (default)
Default Gemini model for reasoning, coding, and everyday work.
Gemini 4 Carbon
Gemini frontier model for complex reasoning and detailed generation.
Gemini 4 Argon
Gemini reasoning model for complex work and detailed generation.
Gemini 3.8 Flash
Latest stable fast tier for chat, analysis, and agentic workflows.
Gemini 3.5 Flash-Lite
Latest stable lightweight tier for concise follow-ups and extraction.
GPT-6.1 Sol
Latest balanced OpenAI model for coding, reasoning, and everyday work.
GPT-6.1 Terra
Current OpenAI model for visual reasoning, writing, and tool-assisted work.
GPT-6 Astra
OpenAI flagship for demanding reasoning and coding tasks.
GPT-6 Luna
Current efficient OpenAI tier for quick, high-volume work.
Claude Opus 5.5
Current Claude tier for long-running coding and knowledge work.
Claude Sonnet 5.5
Balanced Claude model for coding and agent workflows.
Claude Haiku 5.5
Efficient Claude model for quick answers and extraction.
Claude Fable 5.2
Frontier reasoning; requires 30-day provider retention unless exempt.
Earlier model announcements describe their publication context. This list follows the current picker and does not certify provider uptime, account entitlements, or the accuracy of generated output.