MEGA Hub
New
Docs/Get started

Model Catalog#

The Inference Provider catalog lists model and Provider routes that are healthy, validated, and available for new requests. Use it instead of assuming that every model repository or every Provider can serve every task.

Browse live routes#

Open Inference Models to compare one row per model, Provider, and task. The table exposes price, context length, first-token latency, throughput, tool support, and structured-output support.

The same catalog is available without authentication from the Hub API:

BASH
curl https://mega.tensorplay.cn/api/inference/modelscurl https://mega.tensorplay.cn/api/inference/providers/groq/models

The Provider-filtered route returns the same response shape with only that Provider's live mappings. Hub catalog responses may be cached for up to 60 seconds.

The OpenAI-compatible Router exposes an authenticated model list:

BASH
curl https://inference.tensorplay.cn/v1/models \  -H "Authorization: Bearer $MEGA_TOKEN"

Use the Hub catalog when you need MEGA-specific route details. Use /v1/models when an OpenAI-compatible client expects the standard model-list endpoint.

Filter live routes#

Use Inference Models to filter by task or Provider and sort by latency, throughput, and price. For automation, read the Hub catalog JSON or the Provider-filtered endpoint shown above; do not infer availability from a repository tag.

Read a catalog record#

Each model contains a providers array. Important fields are:

Field Meaning
id Public Hub model ID in owner/model form.
provider Provider slug accepted as a model suffix.
task chat-completions, responses, or embeddings.
pricing.input Standard input price in USD per one million tokens.
pricing.output Standard output price in USD per one million tokens.
context_length Validated context window for this mapping.
supports_tools Whether the current validation permits tool calling.
supports_structured_output Whether the current validation permits structured output.
supports_custom_key Whether the Provider can accept a saved custom key.
verified Whether MEGA has verified the Provider route. Organization ownership alone does not set this field.
provider_organization Public organization identity for the route operator, including its independent verified and official status when present.
first_token_latency_ms Best current first-token latency for an available route, when measured.
throughput Best current token throughput for an available route, when measured.

Prices and performance are route properties, not permanent model properties. Compare the route you intend to use and refresh long-lived caches.

Why a route appears or disappears#

A route is published only when all of the following are true:

  • the public model repository exists at the same owner/model ID and its main branch contains README.md;
  • the model mapping is live and has not been deleted;
  • the Provider is enabled and not blocked by risk policy;
  • an eligible route is available, has capacity, and supports the requested task;
  • the mapping supports the requested task.

Repository tags help users discover a model but do not make it callable by themselves. If a route becomes unavailable or incompatible, it is removed from active selection until it can serve the task again. Check the live catalog and Troubleshooting when a route you previously used disappears.

Choose the next step#