Skip to main content

Provider Guides

Loong currently ships 41 built-in provider kinds. This hub exists so the README and chooser pages can stay compact while every built-in provider still has a truthful setup page. Use Providers And Models when you are still choosing a lane. Use this hub when you already know the provider family and want the exact built-in contract, minimal config, and verification path.

Quick Picks

Use this short list when you want a fast starting point before scanning the full matrix below.
If you want…Start hereWhy
the shortest common hosted pathVolcenginethe platform-first hosted lane stays explicit and well-supported in the public docs
another common hosted path with the same shapeBytePlusit keeps the hosted setup path short without hiding the broader provider inventory
a named direct hosted vendor your team already usesOpenAIthe built-in direct hosted contract stays explicit instead of being flattened into one generic story
a coding-dedicated laneBytePlus Coding or Kimi Codingcoding quota and model choice can stay separate from the main assistant lane
a local or self-hosted model serverOllamathe operator-owned server boundary stays visible from the start
a gateway or compatibility endpointOpenRouter or Bedrockroute ownership stays explicit when another layer sits in front of the model vendor

Platform-First Hosted

Common hosted first-run lanes when you want a short reviewed path without hiding the provider family.
ProviderKindDefault auth envModel note
VolcenginevolcengineARK_API_KEYoperator chooses model strategy
BytePlusbyteplusBYTEPLUS_API_KEYoperator chooses model strategy

Direct Hosted Providers

Direct hosted provider families when your team already has a vendor choice and wants it to stay explicit in config.
ProviderKindDefault auth envModel note
OpenAIopenaiOPENAI_API_KEYoperator chooses model strategy
AnthropicanthropicANTHROPIC_API_KEYoperator chooses model strategy
GeminigeminiGEMINI_API_KEYoperator chooses model strategy
DeepSeekdeepseekDEEPSEEK_API_KEYreviewed fallback deepseek-chat
MiniMaxminimaxMINIMAX_API_KEYreviewed fallback MiniMax-M2.7
MistralmistralMISTRAL_API_KEYoperator chooses model strategy
CoherecohereCOHERE_API_KEYoperator chooses model strategy
xAIxaiXAI_API_KEYoperator chooses model strategy
Z.aizaiZAI_API_KEYoperator chooses model strategy
ZhipuzhipuZHIPUAI_API_KEYoperator chooses model strategy
QwenqwenDASHSCOPE_API_KEYoperator chooses model strategy
KimikimiMOONSHOT_API_KEYoperator chooses model strategy
GroqgroqGROQ_API_KEYoperator chooses model strategy
FireworksfireworksFIREWORKS_API_KEYoperator chooses model strategy
TogethertogetherTOGETHER_API_KEYoperator chooses model strategy
PerplexityperplexityPERPLEXITY_API_KEYoperator chooses model strategy
QianfanqianfanQIANFAN_API_KEYoperator chooses model strategy
CerebrascerebrasCEREBRAS_API_KEYoperator chooses model strategy
NVIDIAnvidiaNVIDIA_API_KEYoperator chooses model strategy
SambaNovasambanovaSAMBANOVA_API_KEYoperator chooses model strategy
NovitanovitaNOVITA_API_KEYoperator chooses model strategy
SiliconFlowsiliconflowSILICONFLOW_API_KEYoperator chooses model strategy
StepFunstepfunSTEP_API_KEYoperator chooses model strategy
VeniceveniceVENICE_API_KEYoperator chooses model strategy

Coding-Specialized Providers

Dedicated coding lanes that keep coding quota, pricing, or route ownership separate from the general assistant lane.
ProviderKindDefault auth envModel note
Volcengine Codingvolcengine_codingARK_API_KEYoperator chooses model strategy
BytePlus Codingbyteplus_codingBYTEPLUS_API_KEYoperator chooses model strategy
Kimi Codingkimi_codingKIMI_CODING_API_KEYdefault model kimi-for-coding
Step Planstep_planSTEP_API_KEYoperator chooses model strategy
Bailian Codingbailian_codingBAILIAN_API_KEYoperator chooses model strategy

Gateway And Compatibility Providers

Gateway, proxy, or compatibility endpoints that sit between Loong and upstream model vendors.
ProviderKindDefault auth envModel note
OpenRouteropenrouterOPENROUTER_API_KEYoperator chooses model strategy
Cloudflare AI Gatewaycloudflare_ai_gatewayCLOUDFLARE_API_KEYoperator chooses model strategy
Vercel AI Gatewayvercel_ai_gatewayAI_GATEWAY_API_KEYoperator chooses model strategy
BedrockbedrockAWS_BEARER_TOKEN_BEDROCKoperator chooses model strategy
CustomcustomCUSTOM_PROVIDER_API_KEYoperator chooses model strategy

Local And Self-Hosted Providers

Local or self-hosted servers where the model boundary stays operator-owned.
ProviderKindDefault auth envModel note
Ollamaollamalocal or explicit authoperator chooses model strategy
LM Studiolm_studiolocal or explicit authoperator chooses model strategy
llama.cppllamacpplocal or explicit authoperator chooses model strategy
vLLMvllmlocal or explicit authoperator chooses model strategy
SGLangsglanglocal or explicit authoperator chooses model strategy

Reading Rule

  • Start with Providers And Models when the provider choice is still open.
  • Use the individual guide page when one provider kind is already decided.
  • Keep Provider Recipes for representative rollout paths rather than the full provider inventory.
  • Keep Configuration Patterns open when you need the shared active_provider plus providers.<id> shape.