AI models

Frontier models, assigned per surface.

Run a reasoning model where depth wins and a fast one where a three-second pause reads as hesitation. Same session, no restart.

Per-surface selection

Chat, coding, and system design each get their own model. Set them once in Settings → General.

Reasoning where it counts

System design rewards a model that thinks. Assign GPT-5.4 to that surface alone, or run Claude Opus 4.6 as your Reasoning model.

Speed where it shows

In a behavioral round, latency is visible. Keep something fast on chat so you are never caught waiting.

No API keys to manage

Model access is handled server-side. You never paste a key into the app or babysit a billing dashboard.

Catalog

Every model in the app

Frontier models require a Pro or Pro+ plan. Fast models are available on every plan, including Free.

ModelProviderClassBest for
GPT-5.6 Sol
OpenAIReasoningFrontier-depth reasoning for the hardest rounds
GPT-5.6 Terra
OpenAIFastBalanced depth and latency for everyday rounds
GPT-5.6 Luna
OpenAIFastHigh-volume speed at the lightest cost
GPT-5.5
OpenAIReasoningDeep reasoning across long sessions
GPT-5.5 Pro
OpenAIReasoningThe deepest reasoning, when latency does not matter
GPT-5.4 Pro
OpenAIReasoningExtended reasoning for architecture rounds
GPT-5.4
OpenAIReasoningSystem design and complex technical reasoning
GPT-5.4 Mini
OpenAIFastFast behavioral answers where latency shows
GPT-5.4 Nano
OpenAIFastThe fastest answers at the lightest cost
GPT-5.3 Codex
OpenAICodingLive coding rounds and algorithm problems
GPT-5.2 Pro
OpenAIReasoningExtended reasoning on the 5.2 generation
GPT-5.2
OpenAIFastA balanced general-purpose fallback
GPT-5.1
OpenAIFastA stable older generation, still capable
GPT-4.1 Mini
OpenAIFastLowest latency, lightest credit cost
Claude Opus 4.6
AnthropicReasoningDeep architecture reasoning and nuanced tradeoffs
Claude Sonnet 4.6
AnthropicCodingStrong all-rounder for coding and behavioral
Gemini 2.5 ProSoon
GoogleReasoning
Grok 3Soon
xAIFast

A pairing that works well

Chat · ⌘1
GPT-5.4 Mini

Behavioral answers land fastest when they land immediately.

Coding · ⌘2
GPT-5.3 Codex

Purpose-built for code; strongest on algorithm rounds.

System Design · ⌘3
GPT-5.4

The extra seconds buy you a genuinely better architecture.

Pick your models and start

Fast models on every plan. Frontier models on Pro and Pro+.