Skip to content

Lanes

A lane is a path prefix that selects which catalog and routing policy a request uses. Append the standard OpenAI/Anthropic path after it. All lanes serve both wire protocols (/v1/chat/completions, /v1/responses, /v1/messages).

| Lane | Prefix | Purpose | Status | | --- | --- | --- | --- | | OSS | /oss | Curated open-weight coding flagship. The default. | Stable | | Hugging Face | /hf | Full passthrough of any HF model id. | Stable | | Frontier | /frontier | Claude frontier models. | Stable | | Pro | /pro | Premium tier — higher limits, SLAs. | Coming soon |

The default lane. /oss always serves a curated open-weight catalog (independent of upstream config), so model discovery is reliable for Claude Code Desktop/CLI and OpenAI-compatible clients. Curated aliases are coding-friendly:

| Alias | Family | | --- | --- | | claude-turbo-hub-gpt-oss-20b | GPT-OSS 20B | | claude-turbo-hub-gpt-oss-120b | GPT-OSS 120B | | claude-turbo-hub-qwen3-coder | Qwen3 Coder | | claude-turbo-hub-qwen3-coder-next | Qwen3 Coder Next | | claude-turbo-hub-deepseek-v4-pro | DeepSeek V4 Pro | | claude-turbo-hub-kimi-k2-6 | Kimi K2.6 | | claude-turbo-hub-glm-5-1 | GLM 5.1 |

Always call GET /oss/v1/models for the freshest list. The claude-* alias prefix is deliberate: Claude Code’s model picker only surfaces ids that start with claude/anthropic.

Specify any Hugging Face model id and OpenGateway validates and routes it to an HF provider, across all task types. Use a raw org/model id, optionally with a routing policy suffix:

Terminal window
curl "https://api.opengateway.one/hf/v1/chat/completions" \
-H "Authorization: Bearer $OPENGATEWAY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "Qwen/Qwen3-Coder-480B-A35B-Instruct:fastest",
"messages": [{ "role": "user", "content": "Refactor this function." }]
}'

Passthrough is opt-in by specifying a model id; it never changes the curated default surface. Ids are validated (shape, task↔model match) and upstream identity is always scrubbed from responses and errors.

Claude frontier models on the Anthropic wire (and OpenAI-compat shapes):

| Model ID | Display name | Context | | --- | --- | --- | | turbo-agent-model-claude-opus-4-7 | Claude Opus 4.7 | 1M | | turbo-agent-model-claude-sonnet-4-6 | Claude Sonnet 4.6 | 1M | | turbo-agent-model-claude-haiku-4-5 | Claude Haiku 4.5 | 200k |

Short aliases such as opus-4.7, sonnet-4.6, and haiku-4.5 are also accepted by /frontier. The /frontier lane rejects OSS aliases with 400.

/pro — premium Coming soon

Section titled “/pro — premium ”

A premium tier for higher rate limits, priority routing, and SLAs. The wire protocols and discovery contract are identical to the other lanes, so adopting /pro later is a base-URL change only.

On HF-backed lanes (/oss, /hf) you can pin provider-selection behavior with a suffix on the model id:

| Suffix | Meaning | | --- | --- | | :fastest | Lowest-latency live provider. | | :cheapest | Lowest-cost live provider. | | :preferred | The gateway’s preferred provider for that model. | | :<provider> | Pin a specific provider by name. |

Qwen/Qwen3-Coder-480B-A35B-Instruct:fastest
openai/gpt-oss-120b:cheapest