Providers¶
Auto-generated from brain.providers.provider_registry via sb docs providers --write. Do not hand-edit.
Regenerate with sb docs providers --write (or make docs-reference).
Each provider here is a registered backend the runtime can route to. See How-to → Provider keys for setup.
| Name | Family | Default model | API key env | Model env | Discovery | Self-model |
|---|---|---|---|---|---|---|
anthropic |
anthropic |
claude-opus-4-7 |
ANTHROPIC_API_KEY |
SB_ANTHROPIC_MODEL |
yes | no |
azure |
azure_openai |
— | AZURE_API_KEY |
SB_AZURE_MODEL |
yes | yes |
bedrock |
bedrock |
anthropic.claude-opus-4-7 |
— | SB_BEDROCK_MODEL |
no | yes |
cerebras |
openai_compatible |
gpt-oss-120b |
CEREBRAS_API_KEY |
SB_CEREBRAS_MODEL |
yes | yes |
cohere |
cohere |
command-a-03-2025 |
COHERE_API_KEY |
SB_COHERE_MODEL |
yes | yes |
fireworks |
openai_compatible |
accounts/fireworks/models/gpt-oss-120b |
FIREWORKS_API_KEY |
SB_FIREWORKS_MODEL |
yes | yes |
gemini |
gemini |
gemini-3.1-pro-preview |
GOOGLE_API_KEY |
SB_GEMINI_MODEL |
yes | no |
groq |
openai_compatible |
openai/gpt-oss-120b |
GROQ_API_KEY |
SB_GROQ_MODEL |
yes | yes |
hf |
huggingface |
— | — | SB_HF_MODEL |
no | yes |
huggingface |
huggingface |
— | — | SB_HF_MODEL |
no | yes |
local |
local |
local-echo |
— | SB_LOCAL_MODEL |
no | yes |
mistral |
openai_compatible |
mistral-medium-3-5 |
MISTRAL_API_KEY |
SB_MISTRAL_MODEL |
yes | yes |
nebius |
openai_compatible |
openai/gpt-oss-120b |
NEBIUS_API_KEY |
SB_NEBIUS_MODEL |
yes | yes |
ollama |
ollama |
qwen3.5:9b |
— | SB_OLLAMA_MODEL |
yes | yes |
openai |
openai |
gpt-5.5 |
OPENAI_API_KEY |
SB_OPENAI_MODEL |
yes | no |
openrouter |
openai_compatible |
openai/gpt-5.5 |
OPENROUTER_API_KEY |
SB_OPENROUTER_MODEL |
yes | yes |
parasail |
openai_compatible |
parasail-deepseek-r1 |
PARASAIL_API_KEY |
SB_PARASAIL_MODEL |
yes | yes |
perplexity |
openai_compatible |
sonar-pro |
PERPLEXITY_API_KEY |
SB_PERPLEXITY_MODEL |
no | yes |
qwen |
openai_compatible |
qwen3-coder-plus |
DASHSCOPE_API_KEY |
SB_QWEN_MODEL |
yes | yes |
replicate |
replicate |
openai/gpt-oss-120b |
REPLICATE_API_TOKEN |
SB_REPLICATE_MODEL |
no | yes |
sgl |
openai_compatible |
— | SGL_API_KEY (optional) |
SB_SGL_MODEL |
yes | yes |
vertex |
vertex_openai |
gemini-3.1-pro-preview |
— | SB_VERTEX_MODEL |
no | yes |
vllm |
openai_compatible |
— | VLLM_API_KEY (optional) |
SB_VLLM_MODEL |
yes | yes |
xai |
openai_compatible |
grok-4.3 |
XAI_API_KEY |
SB_XAI_MODEL |
yes | yes |
Field meanings¶
- Family — provider family (e.g.
openai,anthropic,local). - Default model — what the provider uses if no
--modelis passed. - API key env — the environment variable that holds credentials. (optional) providers run without one (typically self-hosted).
- Model env — overrides the default model per-process.
- Discovery —
yesif the provider can list available models programmatically. - Self-model —
yesif the provider manages its own model selection (env vars / local defaults), so passing a tier model name is wrong.