Models Reference

Models Reference

Model Constants

All models are defined as constants in ai/models.go:

// OpenAI
ai.ModelGPT52          // "openai/gpt-5.2"
ai.ModelGPT52Pro        // "openai/gpt-5.2-pro"
ai.ModelGPT51           // "openai/gpt-5.1"
ai.ModelGPT5            // alias of ModelGPT52 (back-compat)
ai.ModelGPT5Mini        // "openai/gpt-5-mini"
ai.ModelGPT5Nano        // "openai/gpt-5-nano"

ai.ModelGPT51Codex      // "openai/gpt-5.1-codex"
ai.ModelGPT51CodexMax   // "openai/gpt-5.1-codex-max"
ai.ModelGPT5Codex       // alias of ModelGPT51CodexMax (back-compat)
ai.ModelGPT5CodexBase   // "openai/gpt-5-codex"

ai.ModelGPT41           // "openai/gpt-4.1"
ai.ModelGPT4o           // "openai/gpt-4o"
ai.ModelGPT4oMini       // "openai/gpt-4o-mini"

ai.ModelO3              // "openai/o3"
ai.ModelO4Mini          // "openai/o4-mini"
ai.ModelO1              // "openai/o1"
ai.ModelO1Mini          // "openai/o1-mini"

// Anthropic
ai.ModelClaudeOpus   // "anthropic/claude-opus-4.5"
ai.ModelClaudeSonnet // "anthropic/claude-sonnet-4.5"
ai.ModelClaudeHaiku  // "anthropic/claude-haiku-4.5"

// Google
// Gemini 3 (preview)
ai.ModelGemini3Pro   // "google/gemini-3-pro-preview"
ai.ModelGemini3Flash // "google/gemini-3-flash-preview"

// Gemini 2.5 (stable)
ai.ModelGemini25Pro       // "google/gemini-2.5-pro"
ai.ModelGemini25Flash     // "google/gemini-2.5-flash"
ai.ModelGemini25FlashLite // "google/gemini-2.5-flash-lite"

// Gemini 2.0 (OpenRouter uses -001 model IDs)
ai.ModelGemini2Flash     // "google/gemini-2.0-flash-001"
ai.ModelGemini2FlashLite // "google/gemini-2.0-flash-lite-001"

// Back-compat alias (maps to Gemini 2.5 Pro)
ai.ModelGemini2Pro // alias of ModelGemini25Pro

// xAI
ai.ModelGrok41Fast // "x-ai/grok-4.1-fast"
ai.ModelGrok3      // "x-ai/grok-3"
ai.ModelGrok3Mini  // "x-ai/grok-3-mini"

// Alibaba
ai.ModelQwen3Next // "qwen/qwen3-next"
ai.ModelQwen3     // "qwen/qwen-3-235b"

// Meta
ai.ModelLlama4 // "meta-llama/llama-4-maverick"

// Mistral
ai.ModelMistralLarge // "mistralai/mistral-large"

Model Shortcuts (Functions)

Convenience functions that return a *Builder:

Function	Model	Notes
ai.GPT5()	openai/gpt-5.2	General purpose, 400K context
ai.GPT5Codex()	openai/gpt-5.1-codex-max	Most intelligent agentic coding model (Codex)
ai.GPT4o()	openai/gpt-4o	Multimodal
ai.GPT4oMini()	openai/gpt-4o-mini	Fast & cheap
ai.O1()	openai/o1	Reasoning model
ai.Claude()	anthropic/claude-opus-4.5	Top coding & safety
ai.ClaudeSonnet()	anthropic/claude-sonnet-4.5	Best balance of intelligence/speed/cost
ai.ClaudeHaiku()	anthropic/claude-haiku-4.5	Fastest Claude (near-frontier)
ai.GeminiPro()	google/gemini-3-pro-preview	Gemini 3 Pro (preview)
ai.Gemini()	google/gemini-3-flash-preview	Gemini 3 Flash (preview)
ai.Gemini25Pro()	google/gemini-2.5-pro	Gemini 2.5 Pro (stable)
ai.Gemini25Flash()	google/gemini-2.5-flash	Gemini 2.5 Flash (stable)
ai.Gemini25FlashLite()	google/gemini-2.5-flash-lite	Gemini 2.5 Flash-Lite (stable)
ai.Gemini2Flash()	google/gemini-2.0-flash-001	Gemini 2.0 Flash (OpenRouter -001)
ai.Gemini2FlashLite()	google/gemini-2.0-flash-lite-001	Gemini 2.0 Flash-Lite (OpenRouter -001)
ai.GrokFast()	x-ai/grok-4.1-fast	2M context, optimized
ai.Grok()	x-ai/grok-3	Real-time knowledge
ai.Qwen()	qwen/qwen3-next	Long context specialist
ai.Llama()	meta-llama/llama-4-maverick	Open weights
ai.Mistral()	mistralai/mistral-large	European AI

Full OpenAI Shortcut List

All OpenAI model constants in models.go also have matching shortcut functions in ai.go:

ai.GPT52()
ai.GPT52Pro()
ai.GPT51()
ai.GPT5Mini()
ai.GPT5Nano()

ai.GPT51Codex()
ai.GPT51CodexMax()
ai.GPT5CodexBase()

ai.GPT5ChatLatest()
ai.ChatGPT4oLatest()

ai.GPT41()

ai.O1Mini()
ai.O1Pro()
ai.O1Preview()
ai.O3()
ai.O3Mini()
ai.O3Pro()
ai.O3DeepResearch()
ai.O4Mini()
ai.O4MiniDeepResearch()

ai.GPTRealtime()
ai.GPTRealtimeMini()
ai.GPTAudio()
ai.GPTAudioMini()

ai.GPT4oMiniTTS()
ai.GPT4oTranscribe()
ai.GPT4oMiniTranscribe()

ai.GPTImage15()
ai.GPTImage1()
ai.GPTImage1Mini()
ai.ChatGPTImageLatest()

ai.GPTOSS120B()
ai.GPTOSS20B()

ai.Sora2()
ai.Sora2Pro()

Using Any Model

For models not in the shortcuts:

// Use any OpenRouter model ID
ai.Use("openai/gpt-4-turbo").Ask("Hello")

// If you're using the OpenRouter provider, bare OpenAI model IDs also work:
ai.Use("gpt-5-mini").Ask("Hello") // normalized to openai/gpt-5-mini

// Or use Model() on existing builder
ai.New(ai.Model("anthropic/claude-3-opus")).Ask("Hello")

Model Info

Get metadata about a model:

info := ai.Models[ai.ModelGPT5]
fmt.Println(info.Name)        // "GPT-5.2"
fmt.Println(info.Provider)    // "OpenAI"
fmt.Println(info.Description) // "General purpose, 400K context"

Switching Models Mid-Builder

builder := ai.GPT5().System("You are helpful")

// Switch to Claude
builder.Model(ai.ModelClaudeOpus).Ask("Hello")

// Or by string
builder.UseModel("anthropic/claude-sonnet-4.5").Ask("Hello")