models

package
v0.704.1 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Sep 8, 2026 License: MIT Imports: 22 Imported by: 0

Documentation

Index

Constants

View Source 
const (
	ProviderAnthropic ModelProvider = "anthropic"

	// Claude 3 family
	Claude3Haiku   ModelID = "anthropic.claude-3-haiku-20240307"
	Claude3Opus    ModelID = "anthropic.claude-3-opus-20240229"
	Claude35Sonnet ModelID = "anthropic.claude-3-5-sonnet-20241022"
	Claude35Haiku  ModelID = "anthropic.claude-3-5-haiku-20241022"
	Claude37Sonnet ModelID = "anthropic.claude-3-7-sonnet-20250219"

	// Claude 4 family
	Claude4Sonnet  ModelID = "anthropic.claude-sonnet-4-20250514"
	Claude4Opus    ModelID = "anthropic.claude-opus-4-20250514"
	Claude4Opus1   ModelID = "anthropic.claude-opus-4-1-20250805"
	Claude45Sonnet ModelID = "anthropic.claude-sonnet-4-5-20250929"
	Claude45Haiku  ModelID = "anthropic.claude-haiku-4-5-20251001"
	Claude45Opus   ModelID = "anthropic.claude-opus-4-5-20251101"
	Claude46Sonnet ModelID = "anthropic.claude-sonnet-4-6"
	Claude46Opus   ModelID = "anthropic.claude-opus-4-6"
)
View Source 
const (
	ProviderCopilot ModelProvider = "copilot"

	// Model IDs match the dynamic format: "copilot.{api-id}"
	CopilotGTP35Turbo      ModelID = "copilot.gpt-3.5-turbo"
	CopilotGPT4o           ModelID = "copilot.gpt-4o"
	CopilotGPT4oMini       ModelID = "copilot.gpt-4o-mini"
	CopilotGPT41           ModelID = "copilot.gpt-4.1"
	CopilotClaude35        ModelID = "copilot.claude-3.5-sonnet"
	CopilotClaude37        ModelID = "copilot.claude-3.7-sonnet"
	CopilotClaude4         ModelID = "copilot.claude-sonnet-4"
	CopilotO1              ModelID = "copilot.o1"
	CopilotO3Mini          ModelID = "copilot.o3-mini"
	CopilotO4Mini          ModelID = "copilot.o4-mini"
	CopilotGemini20        ModelID = "copilot.gemini-2.0-flash"
	CopilotGemini25        ModelID = "copilot.gemini-2.5-pro"
	CopilotGPT4            ModelID = "copilot.gpt-4"
	CopilotClaude37Thought ModelID = "copilot.claude-3.7-sonnet-thought"
	CopilotGPT54           ModelID = "copilot.gpt-5.4"
	CopilotClaudeOpus4     ModelID = "copilot.claude-opus-4"
	CopilotGemini25Flash   ModelID = "copilot.gemini-2.5-flash"
)
View Source 
const (
	CopilotEndpointChatCompletions = "/chat/completions"
	CopilotEndpointResponses       = "/responses"
	CopilotEndpointMessages        = "/v1/messages"
)

Endpoint identifiers as reported by the Copilot /models API in "supported_endpoints".

View Source 
const (
	ProviderAntigravity ModelProvider = "antigravity"
	ProviderGemini      ModelProvider = "gemini"

	// Gemini 3 preview models (require preview access)
	Gemini31ProPreview       ModelID = "gemini.gemini-3.1-pro-preview-customtools"
	Gemini31ProPreviewBase   ModelID = "gemini.gemini-3.1-pro-preview"
	Gemini31FlashLitePreview ModelID = "gemini.gemini-3.1-flash-lite-preview"
	Gemini30ProPreview       ModelID = "gemini.gemini-3-pro-preview"
	Gemini30Flash            ModelID = "gemini.gemini-3-flash-preview"
	// Gemini30Flash legacy alias – config files using "gemini.gemini-3-flash" still resolve.
	Gemini30FlashLegacy ModelID = "gemini.gemini-3-flash"

	// Gemini 2.5 stable models
	Gemini25          ModelID = "gemini.gemini-2.5-pro"
	Gemini25Flash     ModelID = "gemini.gemini-2.5-flash"
	Gemini25FlashLite ModelID = "gemini.gemini-2.5-flash-lite"

	// Gemini 2.5 preview aliases – config files using old preview IDs still resolve.
	Gemini25LegacyPreview      ModelID = "gemini.gemini-2.5-pro-preview-05-06"
	Gemini25FlashLegacyPreview ModelID = "gemini.gemini-2.5-flash-preview-04-17"

	// Gemini 2.0 stable models
	Gemini20Flash     ModelID = "gemini.gemini-2.0-flash"
	Gemini20FlashLite ModelID = "gemini.gemini-2.0-flash-lite"
)
View Source 
const (
	ProviderGROQ ModelProvider = "groq"

	// Model IDs match the dynamic format: "groq.{api-id}"
	QWENQwq                   ModelID = "groq.qwen-qwq-32b"
	Llama4Scout               ModelID = "groq.meta-llama/llama-4-scout-17b-16e-instruct"
	Llama4Maverick            ModelID = "groq.meta-llama/llama-4-maverick-17b-128e-instruct"
	Llama3_3_70BVersatile     ModelID = "groq.llama-3.3-70b-versatile"
	DeepseekR1DistillLlama70b ModelID = "groq.deepseek-r1-distill-llama-70b"
)
View Source 
const (
	ProviderOllama       ModelProvider = "ollama"
	DefaultOllamaBaseURL               = "http://localhost:11434/v1"
	DefaultOllamaRawURL                = "http://localhost:11434"
)
View Source 
const (
	ProviderOpenAI           ModelProvider = "openai"
	ProviderOpenAICompatible ModelProvider = "openai-compatible"

	// Model IDs match the dynamic format: "openai.{api-id}"
	GPT41        ModelID = "openai.gpt-4.1"
	GPT41Mini    ModelID = "openai.gpt-4.1-mini"
	GPT41Nano    ModelID = "openai.gpt-4.1-nano"
	GPT45Preview ModelID = "openai.gpt-4.5-preview"
	GPT4o        ModelID = "openai.gpt-4o"
	GPT4oMini    ModelID = "openai.gpt-4o-mini"
	O1           ModelID = "openai.o1"
	O1Pro        ModelID = "openai.o1-pro"
	O1Mini       ModelID = "openai.o1-mini"
	O3           ModelID = "openai.o3"
	O3Mini       ModelID = "openai.o3-mini"
	O4Mini       ModelID = "openai.o4-mini"
)
View Source 
const (
	ProviderOpenRouter ModelProvider = "openrouter"

	// Model IDs match the dynamic format: "openrouter.{api-id}"
	// OpenRouter API returns IDs in "provider/model" format
	OpenRouterClaude37Sonnet ModelID = "openrouter.anthropic/claude-3.7-sonnet"
	OpenRouterClaude35Haiku  ModelID = "openrouter.anthropic/claude-3.5-haiku"
	OpenRouterGPT41          ModelID = "openrouter.openai/gpt-4.1"
	OpenRouterDeepSeekR1Free ModelID = "openrouter.deepseek/deepseek-r1-0528:free"
)
View Source 
const (
	ProviderVertexAI ModelProvider = "vertexai"

	// Models
	VertexAIGemini25Flash ModelID = "vertexai.gemini-2.5-flash"
	VertexAIGemini25      ModelID = "vertexai.gemini-2.5"
)
View Source 
const (
	ProviderXAI ModelProvider = "xai"

	// Model IDs match the dynamic format: "xai.{api-id}"
	XAIGrok3Beta         ModelID = "xai.grok-3-beta"
	XAIGrok3MiniBeta     ModelID = "xai.grok-3-mini-beta"
	XAIGrok3FastBeta     ModelID = "xai.grok-3-fast-beta"
	XAiGrok3MiniFastBeta ModelID = "xai.grok-3-mini-fast-beta"
)

Variables

View Source 
var AnthropicModels = map[ModelID]Model{

	Claude3Haiku: {
		ID:                  Claude3Haiku,
		Name:                "Claude 3 Haiku",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-3-haiku-20240307",
		CostPer1MIn:         0.25,
		CostPer1MInCached:   0.03,
		CostPer1MOut:        1.25,
		ContextWindow:       200_000,
		DefaultMaxTokens:    4_096,
		SupportsAttachments: true,
	},
	Claude3Opus: {
		ID:                  Claude3Opus,
		Name:                "Claude 3 Opus",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-3-opus-20240229",
		CostPer1MIn:         15.0,
		CostPer1MInCached:   1.50,
		CostPer1MOut:        75.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    4_096,
		SupportsAttachments: true,
	},
	Claude35Sonnet: {
		ID:                  Claude35Sonnet,
		Name:                "Claude 3.5 Sonnet",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-3-5-sonnet-20241022",
		CostPer1MIn:         3.0,
		CostPer1MInCached:   0.30,
		CostPer1MOut:        15.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    8_192,
		SupportsAttachments: true,
	},
	Claude35Haiku: {
		ID:                  Claude35Haiku,
		Name:                "Claude 3.5 Haiku",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-3-5-haiku-20241022",
		CostPer1MIn:         0.80,
		CostPer1MInCached:   0.08,
		CostPer1MOut:        4.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    8_192,
		SupportsAttachments: true,
	},
	Claude37Sonnet: {
		ID:                  Claude37Sonnet,
		Name:                "Claude 3.7 Sonnet",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-3-7-sonnet-20250219",
		CostPer1MIn:         3.0,
		CostPer1MInCached:   0.30,
		CostPer1MOut:        15.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    32_000,
		CanReason:           true,
		SupportsAttachments: true,
	},

	Claude4Sonnet: {
		ID:                  Claude4Sonnet,
		Name:                "Claude Sonnet 4",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-sonnet-4-20250514",
		CostPer1MIn:         3.0,
		CostPer1MInCached:   0.30,
		CostPer1MOut:        15.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    32_000,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Claude4Opus: {
		ID:                  Claude4Opus,
		Name:                "Claude Opus 4",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-opus-4-20250514",
		CostPer1MIn:         15.0,
		CostPer1MInCached:   1.50,
		CostPer1MOut:        75.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    32_000,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Claude4Opus1: {
		ID:                  Claude4Opus1,
		Name:                "Claude Opus 4.1",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-opus-4-1-20250805",
		CostPer1MIn:         15.0,
		CostPer1MInCached:   1.50,
		CostPer1MOut:        75.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    32_000,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Claude45Sonnet: {
		ID:                  Claude45Sonnet,
		Name:                "Claude Sonnet 4.5",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-sonnet-4-5-20250929",
		CostPer1MIn:         3.0,
		CostPer1MInCached:   0.30,
		CostPer1MOut:        15.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    32_000,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Claude45Haiku: {
		ID:                  Claude45Haiku,
		Name:                "Claude Haiku 4.5",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-haiku-4-5-20251001",
		CostPer1MIn:         0.80,
		CostPer1MInCached:   0.08,
		CostPer1MOut:        4.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    32_000,
		SupportsAttachments: true,
	},
	Claude45Opus: {
		ID:                  Claude45Opus,
		Name:                "Claude Opus 4.5",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-opus-4-5-20251101",
		CostPer1MIn:         15.0,
		CostPer1MInCached:   1.50,
		CostPer1MOut:        75.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    32_000,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Claude46Sonnet: {
		ID:                  Claude46Sonnet,
		Name:                "Claude Sonnet 4.6",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-sonnet-4-6",
		CostPer1MIn:         3.0,
		CostPer1MInCached:   0.30,
		CostPer1MOut:        15.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    32_000,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Claude46Opus: {
		ID:                  Claude46Opus,
		Name:                "Claude Opus 4.6",
		Provider:            ProviderAnthropic,
		APIModel:            "claude-opus-4-6",
		CostPer1MIn:         15.0,
		CostPer1MInCached:   1.50,
		CostPer1MOut:        75.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    32_000,
		CanReason:           true,
		SupportsAttachments: true,
	},
}

AnthropicModels lists all supported Anthropic models with their full capabilities. Capabilities are sourced from the Claude Code CLI source (oa() and uM() functions). Context window: 200 000 tokens for all current models. Max output tokens by family (default / upper limit from oa()):

  • Claude 3 Haiku: 4 096 / 4 096
  • Claude 3 Opus: 4 096 / 4 096
  • Claude 3.5 Sonnet / 3.5 Haiku: 8 192 / 8 192
  • Claude 3.7 Sonnet: 32 000 / 64 000
  • Claude 4 Sonnet / Haiku / Opus (all 4.x): 32 000 / 64 000
View Source 
var AntigravityModels = map[ModelID]Model{
	AntigravityGemini30Pro: {
		ID:                  AntigravityGemini30Pro,
		Name:                "Gemini 3 Pro",
		Provider:            ProviderAntigravity,
		APIModel:            "gemini-3-pro-preview",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		CanReason:           true,
		SupportsAttachments: true,
	},
	AntigravityGemini31Pro: {
		ID:                  AntigravityGemini31Pro,
		Name:                "Gemini 3.1 Pro",
		Provider:            ProviderAntigravity,
		APIModel:            "gemini-3.1-pro-preview-customtools",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		CanReason:           true,
		SupportsAttachments: true,
	},
	AntigravityGemini30Flash: {
		ID:                  AntigravityGemini30Flash,
		Name:                "Gemini 3 Flash",
		Provider:            ProviderAntigravity,
		APIModel:            "gemini-3-flash-preview",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		SupportsAttachments: true,
	},
}
View Source 
var AzureModels = map[ModelID]Model{
	AzureGPT41: {
		ID:                  AzureGPT41,
		Name:                "Azure OpenAI – GPT 4.1",
		Provider:            ProviderAzure,
		APIModel:            "gpt-4.1",
		CostPer1MIn:         2.00,
		CostPer1MInCached:   0.50,
		CostPer1MOut:        8.00,
		CostPer1MOutCached:  0.0,
		ContextWindow:       1_047_576,
		DefaultMaxTokens:    20000,
		SupportsAttachments: true,
	},
	AzureGPT41Mini: {
		ID:                  AzureGPT41Mini,
		Name:                "Azure OpenAI – GPT 4.1 mini",
		Provider:            ProviderAzure,
		APIModel:            "gpt-4.1-mini",
		CostPer1MIn:         0.40,
		CostPer1MInCached:   0.10,
		CostPer1MOut:        1.60,
		CostPer1MOutCached:  0.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    20000,
		SupportsAttachments: true,
	},
	AzureGPT41Nano: {
		ID:                  AzureGPT41Nano,
		Name:                "Azure OpenAI – GPT 4.1 nano",
		Provider:            ProviderAzure,
		APIModel:            "gpt-4.1-nano",
		CostPer1MIn:         0.10,
		CostPer1MInCached:   0.025,
		CostPer1MOut:        0.40,
		CostPer1MOutCached:  0.0,
		ContextWindow:       1_047_576,
		DefaultMaxTokens:    20000,
		SupportsAttachments: true,
	},
	AzureGPT45Preview: {
		ID:                  AzureGPT45Preview,
		Name:                "Azure OpenAI – GPT 4.5 preview",
		Provider:            ProviderAzure,
		APIModel:            "gpt-4.5-preview",
		CostPer1MIn:         75.00,
		CostPer1MInCached:   37.50,
		CostPer1MOut:        150.00,
		CostPer1MOutCached:  0.0,
		ContextWindow:       128_000,
		DefaultMaxTokens:    15000,
		SupportsAttachments: true,
	},
	AzureGPT4o: {
		ID:                  AzureGPT4o,
		Name:                "Azure OpenAI – GPT-4o",
		Provider:            ProviderAzure,
		APIModel:            "gpt-4o",
		CostPer1MIn:         2.50,
		CostPer1MInCached:   1.25,
		CostPer1MOut:        10.00,
		CostPer1MOutCached:  0.0,
		ContextWindow:       128_000,
		DefaultMaxTokens:    4096,
		SupportsAttachments: true,
	},
	AzureGPT4oMini: {
		ID:                  AzureGPT4oMini,
		Name:                "Azure OpenAI – GPT-4o mini",
		Provider:            ProviderAzure,
		APIModel:            "gpt-4o-mini",
		CostPer1MIn:         0.15,
		CostPer1MInCached:   0.075,
		CostPer1MOut:        0.60,
		CostPer1MOutCached:  0.0,
		ContextWindow:       128_000,
		SupportsAttachments: true,
	},
	AzureO1: {
		ID:                  AzureO1,
		Name:                "Azure OpenAI – O1",
		Provider:            ProviderAzure,
		APIModel:            "o1",
		CostPer1MIn:         15.00,
		CostPer1MInCached:   7.50,
		CostPer1MOut:        60.00,
		CostPer1MOutCached:  0.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    50000,
		CanReason:           true,
		SupportsAttachments: true,
	},
	AzureO1Mini: {
		ID:                  AzureO1Mini,
		Name:                "Azure OpenAI – O1 mini",
		Provider:            ProviderAzure,
		APIModel:            "o1-mini",
		CostPer1MIn:         1.10,
		CostPer1MInCached:   0.55,
		CostPer1MOut:        4.40,
		CostPer1MOutCached:  0.0,
		ContextWindow:       128_000,
		DefaultMaxTokens:    50000,
		CanReason:           true,
		SupportsAttachments: true,
	},
	AzureO3: {
		ID:                  AzureO3,
		Name:                "Azure OpenAI – O3",
		Provider:            ProviderAzure,
		APIModel:            "o3",
		CostPer1MIn:         10.00,
		CostPer1MInCached:   2.50,
		CostPer1MOut:        40.00,
		CostPer1MOutCached:  0.0,
		ContextWindow:       200_000,
		CanReason:           true,
		SupportsAttachments: true,
	},
	AzureO3Mini: {
		ID:                  AzureO3Mini,
		Name:                "Azure OpenAI – O3 mini",
		Provider:            ProviderAzure,
		APIModel:            "o3-mini",
		CostPer1MIn:         1.10,
		CostPer1MInCached:   0.55,
		CostPer1MOut:        4.40,
		CostPer1MOutCached:  0.0,
		ContextWindow:       200_000,
		DefaultMaxTokens:    50000,
		CanReason:           true,
		SupportsAttachments: false,
	},
	AzureO4Mini: {
		ID:                  AzureO4Mini,
		Name:                "Azure OpenAI – O4 mini",
		Provider:            ProviderAzure,
		APIModel:            "o4-mini",
		CostPer1MIn:         1.10,
		CostPer1MInCached:   0.275,
		CostPer1MOut:        4.40,
		CostPer1MOutCached:  0.0,
		ContextWindow:       128_000,
		DefaultMaxTokens:    50000,
		CanReason:           true,
		SupportsAttachments: true,
	},
}
View Source 
var GeminiModels = map[ModelID]Model{

	Gemini31ProPreview: {
		ID:                  Gemini31ProPreview,
		Name:                "Gemini 3.1 Pro Preview (Custom Tools)",
		Provider:            ProviderGemini,
		APIModel:            "gemini-3.1-pro-preview-customtools",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Gemini31ProPreviewBase: {
		ID:                  Gemini31ProPreviewBase,
		Name:                "Gemini 3.1 Pro Preview",
		Provider:            ProviderGemini,
		APIModel:            "gemini-3.1-pro-preview",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Gemini31FlashLitePreview: {
		ID:                  Gemini31FlashLitePreview,
		Name:                "Gemini 3.1 Flash Lite Preview",
		Provider:            ProviderGemini,
		APIModel:            "gemini-3.1-flash-lite-preview",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Gemini30ProPreview: {
		ID:                  Gemini30ProPreview,
		Name:                "Gemini 3 Pro Preview",
		Provider:            ProviderGemini,
		APIModel:            "gemini-3-pro-preview",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Gemini30Flash: {
		ID:                  Gemini30Flash,
		Name:                "Gemini 3 Flash Preview",
		Provider:            ProviderGemini,
		APIModel:            "gemini-3-flash-preview",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		CanReason:           true,
		SupportsAttachments: true,
	},

	Gemini30FlashLegacy: {
		ID:                  Gemini30FlashLegacy,
		Name:                "Gemini 3 Flash Preview",
		Provider:            ProviderGemini,
		APIModel:            "gemini-3-flash-preview",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		CanReason:           true,
		SupportsAttachments: true,
	},

	Gemini25: {
		ID:                  Gemini25,
		Name:                "Gemini 2.5 Pro",
		Provider:            ProviderGemini,
		APIModel:            "gemini-2.5-pro",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Gemini25Flash: {
		ID:                  Gemini25Flash,
		Name:                "Gemini 2.5 Flash",
		Provider:            ProviderGemini,
		APIModel:            "gemini-2.5-flash",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Gemini25FlashLite: {
		ID:                  Gemini25FlashLite,
		Name:                "Gemini 2.5 Flash Lite",
		Provider:            ProviderGemini,
		APIModel:            "gemini-2.5-flash-lite",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		SupportsAttachments: true,
	},

	Gemini25LegacyPreview: {
		ID:                  Gemini25LegacyPreview,
		Name:                "Gemini 2.5 Pro",
		Provider:            ProviderGemini,
		APIModel:            "gemini-2.5-pro",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		CanReason:           true,
		SupportsAttachments: true,
	},
	Gemini25FlashLegacyPreview: {
		ID:                  Gemini25FlashLegacyPreview,
		Name:                "Gemini 2.5 Flash",
		Provider:            ProviderGemini,
		APIModel:            "gemini-2.5-flash",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		CanReason:           true,
		SupportsAttachments: true,
	},

	Gemini20Flash: {
		ID:                  Gemini20Flash,
		Name:                "Gemini 2.0 Flash",
		Provider:            ProviderGemini,
		APIModel:            "gemini-2.0-flash",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		SupportsAttachments: true,
	},
	Gemini20FlashLite: {
		ID:                  Gemini20FlashLite,
		Name:                "Gemini 2.0 Flash Lite",
		Provider:            ProviderGemini,
		APIModel:            "gemini-2.0-flash-lite",
		ContextWindow:       1_000_000,
		DefaultMaxTokens:    8_192,
		SupportsAttachments: true,
	},
}
View Source 
var ProviderPopularity = map[ModelProvider]int{
	ProviderCopilot:          1,
	ProviderAnthropic:        2,
	ProviderOpenAI:           3,
	ProviderAntigravity:      4,
	ProviderOllama:           5,
	ProviderLlamaCpp:         6,
	ProviderGemini:           7,
	ProviderGROQ:             8,
	ProviderOpenRouter:       9,
	ProviderBedrock:          10,
	ProviderAzure:            11,
	ProviderVertexAI:         12,
	ProviderOpenAICompatible: 13,
}

Providers in order of popularity

View Source 
var VertexAIGeminiModels = map[ModelID]Model{
	VertexAIGemini25Flash: {
		ID:                  VertexAIGemini25Flash,
		Name:                "VertexAI: Gemini 2.5 Flash",
		Provider:            ProviderVertexAI,
		APIModel:            "gemini-2.5-flash-preview-04-17",
		CostPer1MIn:         0.15,
		CostPer1MInCached:   0,
		CostPer1MOut:        0.60,
		CostPer1MOutCached:  0,
		ContextWindow:       1000000,
		DefaultMaxTokens:    50000,
		SupportsAttachments: true,
	},
	VertexAIGemini25: {
		ID:                  VertexAIGemini25,
		Name:                "VertexAI: Gemini 2.5 Pro",
		Provider:            ProviderVertexAI,
		APIModel:            "gemini-2.5-pro-preview-03-25",
		CostPer1MIn:         1.25,
		CostPer1MInCached:   0,
		CostPer1MOut:        10,
		CostPer1MOutCached:  0,
		ContextWindow:       1000000,
		DefaultMaxTokens:    50000,
		SupportsAttachments: true,
	},
}

Functions

func CopilotModelUsesResponsesAPI added in v0.624.3 

func CopilotModelUsesResponsesAPI(m Model) bool

CopilotModelUsesResponsesAPI reports whether a model must be driven through the OpenAI Responses API (/responses) instead of Chat Completions.

The Copilot /models API advertises the routes each model accepts, and that metadata is authoritative: models such as "mai-code-1-flash-picker" or "gpt-5.6-luna" only accept /responses while carrying names that no gpt-version heuristic can classify. This mirrors how the official VS Code extension routes requests (see chatEndpoint.ts, useResponsesApi).

When a model advertises both routes we keep the historical name-based decision, so models like gpt-5-mini keep working over Chat Completions. Models with no advertised endpoints (static catalogue entries, older API responses) fall back to the name heuristic entirely.

func DefaultReasoningEffort added in v0.646.4 

func DefaultReasoningEffort(model Model) string

DefaultReasoningEffort returns a reasoning-effort value that is always valid for the model: "medium" when acceptable, else "high" (the most common safe strong default), else the weakest value the model supports. It never returns a value absent from the model's allowed set; for a model with no known set it falls back to "medium" to preserve prior behaviour.

func DeleteSupportedModels added in v0.644.0 

func DeleteSupportedModels(ids ...ModelID)

DeleteSupportedModels removes models from the catalogue.

func EnrichModelFromModelsDev added in v0.629.0 

func EnrichModelFromModelsDev(ctx context.Context, model *Model)

EnrichModelFromModelsDev fills in metadata the provider's own listing API did not report — most importantly per-token costs, which almost no provider returns and without which the session cost panel stays empty.

It only ever writes fields that are currently zero/empty: whatever the provider (or Pando's curated catalog) already stated stays authoritative, so a stale or wrong catalog entry can never downgrade known-good metadata. Boolean capability flags are only raised, never cleared, for the same reason.

func EnrichRegisteredModels added in v0.629.0 

func EnrichRegisteredModels(ctx context.Context)

EnrichRegisteredModels tops up every model already in the registry with the catalog data it is missing. Curated static entries (Antigravity, Bedrock, Vertex AI, …) are hand-written without prices, so without this pass they would keep reporting a zero session cost even though the catalog knows their rates.

Safe to call repeatedly; it is a no-op when the catalog is unavailable.

func GetAllModels added in v0.2.0 

func GetAllModels() map[ModelID]Model

GetAllModels returns both static and dynamic models

func IsCopilotAnthropicModel added in v0.3.0 

func IsCopilotAnthropicModel(apiModel string) bool

IsCopilotAnthropicModel returns true if the given API model name is an Anthropic model served through the GitHub Copilot API (detected by name containing "claude").

func IsCopilotResponsesAPIModel added in v0.100.0 

func IsCopilotResponsesAPIModel(apiModel string) bool

IsCopilotResponsesAPIModel reports whether a model requires the Responses API based on its name alone. It is the fallback used by CopilotModelUsesResponsesAPI when the provider does not advertise the model's supported endpoints; prefer that function whenever a full Model is available. GPT-5+ models (except gpt-5-mini variants) require the Responses API.

func LoadModelCache added in v0.3.0 

func LoadModelCache() error

LoadModelCache loads previously cached dynamic models into SupportedModels(). Cached models do not overwrite static entries (Azure, VertexAI, Bedrock).

func ModelsDevMetadata added in v0.629.0 

func ModelsDevMetadata(ctx context.Context, provider ModelProvider, apiModel string) (modelsdev.Model, bool)

ModelsDevMetadata returns the models.dev entry describing a provider/model pair. ok is false whenever the catalog is unavailable or the model is unknown, which every caller must treat as "keep what you already have".

func NormalizeReasoningEffort added in v0.646.4 

func NormalizeReasoningEffort(model Model, value string) string

NormalizeReasoningEffort returns the canonical form of value when it is one of the model's accepted reasoning efforts, or "" otherwise. Passing an empty value also returns "" (callers distinguish "unset" from "invalid").

func ProviderSupportsModelListing added in v0.503.5 

func ProviderSupportsModelListing(provider ModelProvider) bool

ProviderSupportsModelListing reports whether a provider exposes a model-listing API that FetchModelsFromProvider can query. Providers that return false (e.g. Azure, Vertex AI, Bedrock) rely on a curated static catalog instead, and must register per-account model copies separately to support multiple accounts.

func ReasoningEffortsFor added in v0.646.4 

func ReasoningEffortsFor(model Model) []string

ReasoningEffortsFor returns the ordered reasoning-effort values a model accepts. A non-empty list on Model.ReasoningEfforts (from the models.dev catalog) is authoritative; otherwise a per-provider/family fallback table is consulted. It returns nil when the model's effort set is unknown.

func RefreshProviderModels added in v0.2.0 

func RefreshProviderModels(ctx context.Context, provider ModelProvider, apiKey string, bearerToken string, baseURL string) error

RefreshProviderModels fetches and registers models from a provider. Models previously registered for this provider that are no longer returned by the API are removed from the registry and cache.

func RefreshProviderModelsForAccount added in v0.252.0 

func RefreshProviderModelsForAccount(ctx context.Context, params AccountModelRefreshParams) error

RefreshProviderModelsForAccount fetches and registers models for a named provider account. Model IDs are prefixed with accountID when AllAccountsOfType > 1 (disambiguates multiple accounts of same type). Models previously registered for this account that are no longer returned by the API are removed.

func RegisterDynamicModel added in v0.2.0 

func RegisterDynamicModel(model Model)

RegisterDynamicModel adds a dynamically discovered model.

func RegisterDynamicModels added in v0.644.0 

func RegisterDynamicModels(batch []Model)

RegisterDynamicModels adds a batch of dynamically discovered models with a single catalogue update, avoiding one copy-on-write per model.

func ResolveLlamaCppBaseURL added in v0.291.0 

func ResolveLlamaCppBaseURL(configured string) string

ResolveLlamaCppBaseURL returns the llama-server base URL, consulting the LLAMACPP_BASE_URL environment variable as a fallback before using the built-in default (http://localhost:8080/v1).

func ResolveOllamaBaseURL added in v0.2.0 

func ResolveOllamaBaseURL(configured string) string

func ResolveOllamaRawBaseURL added in v0.2.0 

func ResolveOllamaRawBaseURL(configured string) string

ResolveOllamaRawBaseURL returns the raw Ollama base URL without the /v1 suffix, suitable for calling native Ollama endpoints like /api/tags.

func SaveModelCache added in v0.3.0 

func SaveModelCache() error

SaveModelCache saves dynamically discovered models to a local cache file. This allows models to be available immediately on the next startup.

func SetSupportedModel added in v0.644.0 

func SetSupportedModel(model Model)

SetSupportedModel adds or replaces a single model in the catalogue.

func SetSupportedModels added in v0.644.0 

func SetSupportedModels(batch map[ModelID]Model)

SetSupportedModels adds or replaces several models in a single update, which is what callers registering a batch should use: the catalogue is copy-on-write and a per-model write would copy it once per entry.

func SupportedModels 

func SupportedModels() map[ModelID]Model

SupportedModels returns the current model catalogue. The returned map is shared and MUST NOT be modified; use the registration helpers instead.

Types

type AccountModelRefreshParams added in v0.252.0 

type AccountModelRefreshParams struct {
	AccountID    string
	ProviderType ModelProvider
	APIKey       string
	BearerToken  string
	BaseURL      string
	ExtraHeaders map[string]string
	// AllAccountsOfType is the count of non-disabled accounts sharing ProviderType,
	// used to decide whether to prefix the model ID with AccountID.
	AllAccountsOfType int
}

AccountModelRefreshParams holds the parameters needed to refresh models for a named account.

type FetchedModel added in v0.2.0 

type FetchedModel struct {
	ID              string `json:"id"`
	Name            string `json:"name,omitempty"`
	Description     string `json:"description,omitempty"`
	Created         int64  `json:"created,omitempty"`
	ContextWindow   int64  `json:"context_window,omitempty"`
	MaxOutputTokens int64  `json:"max_output_tokens,omitempty"`
	// SupportedEndpoints lists the API routes the provider accepts for this
	// model (e.g. "/responses", "/chat/completions"). Empty when the provider's
	// listing API does not report it, in which case callers fall back to
	// name-based heuristics.
	SupportedEndpoints      []string `json:"supported_endpoints,omitempty"`
	CanReason               bool     `json:"can_reason,omitempty"`
	SupportsReasoningEffort bool     `json:"supports_reasoning_effort,omitempty"`
	SupportsAttachments     bool     `json:"supports_attachments,omitempty"`
}

FetchedModel represents a model returned by a provider's API

func FetchModelsFromProvider added in v0.2.0 

func FetchModelsFromProvider(ctx context.Context, provider ModelProvider, apiKey string, bearerToken string, baseURL string) ([]FetchedModel, error)

FetchModelsFromProvider queries a provider's API for available models

type Model 

type Model struct {
	ID                      ModelID       `json:"id"`
	Name                    string        `json:"name"`
	Provider                ModelProvider `json:"provider"`
	APIModel                string        `json:"api_model"`
	CostPer1MIn             float64       `json:"cost_per_1m_in"`
	CostPer1MOut            float64       `json:"cost_per_1m_out"`
	CostPer1MInCached       float64       `json:"cost_per_1m_in_cached"`
	CostPer1MOutCached      float64       `json:"cost_per_1m_out_cached"`
	ContextWindow           int64         `json:"context_window"`
	DefaultMaxTokens        int64         `json:"default_max_tokens"`
	CanReason               bool          `json:"can_reason"`
	SupportsReasoningEffort bool          `json:"supports_reasoning_effort"`
	// ReasoningEfforts lists the ordered reasoning-effort values (weakest to
	// strongest) the model accepts for its effort-style control, e.g. OpenAI
	// "reasoning_effort" or Anthropic adaptive "thinking" budget. It comes from
	// the models.dev catalog (ReasoningOptions) when available. Empty means
	// "unknown", and the resolver falls back to a per-family table.
	ReasoningEfforts    []string `json:"reasoning_efforts,omitempty"`
	SupportsAttachments bool     `json:"supports_attachments"`
	// SupportedEndpoints lists the API routes the provider accepts for this
	// model, as reported by its model-listing API (e.g. "/responses",
	// "/chat/completions"). Empty for statically-defined models and for
	// providers that do not report it.
	SupportedEndpoints []string `json:"supported_endpoints,omitempty"`
	// AccountID is the ProviderAccount.ID that this model belongs to.
	// Empty means the model comes from the legacy single-account system.
	AccountID string `json:"account_id,omitempty"`
	// Description, Knowledge (training cutoff, e.g. "2026-01") and ReleaseDate
	// are descriptive fields shown in the model selectors. They are usually
	// absent from provider listing APIs and come from the models.dev catalog.
	Description string `json:"description,omitempty"`
	Knowledge   string `json:"knowledge,omitempty"`
	ReleaseDate string `json:"release_date,omitempty"`
	// ImageMaxLongSidePx is the model's vision long-side pixel limit. Images
	// larger than this are rescaled by the provider before encoding, so Pando
	// resizes to this value locally to save bandwidth. 0 means "unknown": use
	// ImageLongSideLimit() which falls back to a per-family default.
	ImageMaxLongSidePx int `json:"image_max_long_side_px,omitempty"`
}

func AccountScopedStaticModels added in v0.503.5 

func AccountScopedStaticModels(provider ModelProvider, accountID string, allAccountsOfType int) []Model

AccountScopedStaticModels returns per-account copies of a provider's static models, with account-prefixed IDs and AccountID set. It returns nothing when fewer than two accounts of the type exist: with a single account the prefix equals the provider type, so the prefixed ID would collide with the static one and the account-less static models (which already resolve to that single account) are used as-is. Used both to register the models (RefreshProviderModels- ForAccount) and to list them (the /api/v1/models handler) for providers that have no model-listing API.

func StaticModelsForProvider added in v0.503.5 

func StaticModelsForProvider(provider ModelProvider) []Model

StaticModelsForProvider returns copies of every statically-defined (account-less) model for a provider type (exported for listing handlers).

func (Model) DisplayLabel added in v0.252.0 

func (m Model) DisplayLabel(sameTypeAccountCount int) string

DisplayLabel returns the display label for a model. sameTypeAccountCount is how many non-disabled accounts share this model's provider type. If > 1, the account ID is prefixed to disambiguate.

func (Model) ImageLongSideLimit added in v0.629.4 

func (m Model) ImageLongSideLimit() int

ImageLongSideLimit returns the long-side pixel cap to resize images to before sending them to this model. Prefers the explicit ImageMaxLongSidePx (e.g. from the models.dev catalog), otherwise infers a per-family default.

type ModelID 

type ModelID string
const (
	AntigravityGemini30Pro   ModelID = "antigravity/gemini-3-pro"
	AntigravityGemini31Pro   ModelID = "antigravity/gemini-3.1-pro"
	AntigravityGemini30Flash ModelID = "antigravity/gemini-3-flash"
)
const (
	AzureGPT41        ModelID = "azure.gpt-4.1"
	AzureGPT41Mini    ModelID = "azure.gpt-4.1-mini"
	AzureGPT41Nano    ModelID = "azure.gpt-4.1-nano"
	AzureGPT45Preview ModelID = "azure.gpt-4.5-preview"
	AzureGPT4o        ModelID = "azure.gpt-4o"
	AzureGPT4oMini    ModelID = "azure.gpt-4o-mini"
	AzureO1           ModelID = "azure.o1"
	AzureO1Mini       ModelID = "azure.o1-mini"
	AzureO3           ModelID = "azure.o3"
	AzureO3Mini       ModelID = "azure.o3-mini"
	AzureO4Mini       ModelID = "azure.o4-mini"
)
const (
	// Bedrock
	BedrockClaude37Sonnet ModelID = "bedrock.claude-3.7-sonnet"
)

Model IDs

func CanonicalAccountModelID added in v0.503.5 

func CanonicalAccountModelID(providerType ModelProvider, accountID string, allAccountsOfType int, apiModelID string) ModelID

CanonicalAccountModelID returns the registry model ID for a model fetched from a provider account, matching exactly the IDs produced by RefreshProviderModelsForAccount (i.e. "<prefix>.<apiModelID>"). Callers that list models for selection must use this so the IDs they expose are the same ones the rest of the system (model cache, agent validation, TUI) recognises. Using a bare, non-prefixed ID causes validateAgent to reject the saved agent model on the next config reload and silently revert it to a provider default.

func NormalizeModelID added in v0.210.0 

func NormalizeModelID(input string) ModelID

func ResolveModelID added in v0.503.5 

func ResolveModelID(input ModelID) (ModelID, bool)

ResolveModelID maps a possibly non-canonical model ID to a registered model ID. It returns (id, true) when the input is already registered, or when exactly one registered model matches it by APIModel or by a "<provider>.<input>" suffix. Otherwise it returns (input, false). This lets callers repair legacy/bare model IDs (e.g. "gpt-5.4-mini") to their canonical form ("copilot.gpt-5.4-mini") instead of discarding the user's selection.

type ModelProvider 

type ModelProvider string
const (
	ProviderLlamaCpp       ModelProvider = "llama-cpp"
	DefaultLlamaCppBaseURL               = "http://localhost:8080/v1"
)
const (
	ProviderBedrock ModelProvider = "bedrock"
	// ForTests
	ProviderMock ModelProvider = "__mock"
)
const ProviderAzure ModelProvider = "azure"
const (
	ProviderLocal ModelProvider = "local"
)

Source Files

View all Source files

Directories

Path Synopsis
Package modelsdev provides read-only access to the models.dev catalog (https://github.com/anomalyco/models.dev), a community-maintained database of model pricing and capabilities per provider.
 Package modelsdev provides read-only access to the models.dev catalog (https://github.com/anomalyco/models.dev), a community-maintained database of model pricing and capabilities per provider.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.