Providers

GoAI SDK supports 25+ LLM providers with a single unified API. Most hosted providers auto-resolve credentials from environment variables; local/custom providers (for example ollama, vllm, compat) typically use explicit options.

The core functions — GenerateText, StreamText, GenerateObject[T], StreamObject[T], Embed, EmbedMany, GenerateImage, and GenerateVideo — share one API, but provider capabilities vary.

Tier 1

Dedicated implementations with extended API support.

Provider	Chat	Embed	Image	Provider Tools	Auth Env Var
OpenAI	✅ gpt-4o, o3	✅ text-embedding-3-*	✅ gpt-image-1	4 tools	OPENAI_API_KEY
Anthropic	✅ Anthropic models	-	-	14 tools	ANTHROPIC_API_KEY
Google	✅ gemini-*	✅ text-embedding-004	✅ imagen-*	5 tools	GOOGLE_GENERATIVE_AI_API_KEY or GEMINI_API_KEY
Bedrock	✅ anthropic.*, meta.*	✅ titan-embed-*, cohere.embed-*	—	—	AWS_ACCESS_KEY_ID
Azure	✅ gpt-4o, claude-*	—	✅	—	AZURE_OPENAI_API_KEY
Vertex AI	✅ gemini-*, claude-*	✅	✅	—	ADC/OAuth; API-key fallback applies only outside native Gemini mode

Tier 2

Provider	Chat	Embed	Image	Provider Tools	Auth Env Var
Cohere	✅ command-r-*	✅ embed-*	—	—	COHERE_API_KEY
Mistral	✅ mistral-large, magistral-*	—	—	—	MISTRAL_API_KEY
xAI (Grok)	✅ grok-*	—	—	2 tools	XAI_API_KEY
Groq	✅ llama-*, mixtral-*	—	—	1 tool	GROQ_API_KEY
DeepSeek	✅ deepseek-chat, deepseek-reasoner	—	—	—	DEEPSEEK_API_KEY
MiniMax	✅ MiniMax-M2.7, MiniMax-M2.5, MiniMax-M2.1, MiniMax-M2	—	—	—	MINIMAX_API_KEY

Tier 3

Most use the shared internal/openaicompat codec (some wrappers delegate via provider/compat).

Provider	Endpoint	Auth Env Var
Fireworks	api.fireworks.ai	FIREWORKS_API_KEY
Together	api.together.ai	TOGETHER_AI_API_KEY (or TOGETHER_API_KEY)
DeepInfra	api.deepinfra.com	DEEPINFRA_API_KEY
OpenRouter	openrouter.ai	OPENROUTER_API_KEY
Requesty	router.requesty.ai	REQUESTY_API_KEY
Perplexity	api.perplexity.ai	PERPLEXITY_API_KEY
Cerebras	api.cerebras.ai	CEREBRAS_API_KEY
RunPod	api.runpod.ai	RUNPOD_API_KEY
Cloudflare	api.cloudflare.com	CLOUDFLARE_API_TOKEN, CLOUDFLARE_ACCOUNT_ID
FPT Cloud	mkp-api.fptcloud.com / .jp	FPT_API_KEY, FPT_REGION
NVIDIA NIM	integrate.api.nvidia.com/v1	NVIDIA_API_KEY, NVIDIA_BASE_URL
Heabsy	api.heabsy.com	HEABSY_API_KEY, HEABSY_BASE_URL
Opper	api.opper.ai	OPPER_API_KEY, OPPER_BASE_URL
Runcrate	api.runcrate.ai	RUNCRATE_API_KEY, RUNCRATE_BASE_URL
Cheaper Inference	api.cheaperinference.com/v1	CHEAPER_INFERENCE_API_KEY, CHEAPER_INFERENCE_BASE_URL

Local / Custom

Provider	Default Endpoint	Auth	Features
Ollama	localhost:11434	None required	Embedding support
llama.cpp (provider/llamacpp)	localhost:8080	Optional	Chat and embedding support
vLLM	localhost:8000/v1	Optional	Embedding support
Generic Compatible	(required)	Configurable	Any OpenAI-compatible endpoint

Common Options

Most providers support these options (Bedrock uses AWS credential options; Azure uses WithEndpoint; Ollama requires no auth):

openai.WithAPIKey(key)        // Static API key
openai.WithTokenSource(ts)    // Dynamic auth (OAuth, service accounts)
openai.WithBaseURL(url)       // Override endpoint
openai.WithHeaders(h)         // Custom HTTP headers
openai.WithHTTPClient(c)      // Custom HTTP transport

Each provider package exports its own With* functions with broadly similar signatures (with provider-specific exceptions where needed).