NEAR AI Cloud brings GLM 5.3 Flash to OpenRouter

NEAR AI Cloud became an OpenRouter provider on 28 September 2026, giving developers another way to run Z.ai’s GLM 5.3 Flash model. OpenRouter’s NEAR AI listing shows one model. People already using OpenRouter can select the provider with their existing API key and credits, according to NEAR’s announcement.

NEAR AI runs the inference service; Z.ai developed the model. The integration puts NEAR’s model endpoint inside OpenRouter’s catalogue, where one API can route requests among providers. It also gives developers two distinct routes to consider: OpenRouter for shared access and provider switching, or NEAR’s direct API for its confidential inference service.

One model is available through NEAR AI

GLM 5.3 Flash is the only model on NEAR AI’s OpenRouter provider page as of 30 September 2026. It accepts text, images and video as input and returns text, according to OpenRouter’s model page. The listed context window is 1,048,576 tokens, large enough for substantial documents or codebases, subject to the limits of the chosen endpoint and request.

The OpenRouter listing displayed the promotional rate when checked on 30 September; the NEAR model page lists the standard rate and a $0.035 rate per million cached input tokens. Provider prices and promotions can change, so the live listing is the place to check before sending a large workload.

You can put NEAR AI first in the route

The near-ai slug tells OpenRouter to try NEAR AI before other providers of GLM 5.3 Flash. NEAR gives this request body in its launch instructions:

{
  "model": "z-ai/glm-5.3-flash",
  "messages": [{ "role": "user", "content": "Hello" }],
  "provider": { "order": ["near-ai"] }
}

An application sends that body to OpenRouter’s chat-completions endpoint with its OpenRouter key. The order field makes NEAR AI the first choice. Fallbacks remain enabled by default, so OpenRouter may send the request to another provider if NEAR AI cannot serve it. To require NEAR AI for that request, add "allow_fallbacks": false inside provider; the call then fails when the selected route is unavailable. The OpenRouter routing guide describes the same preference controls. A separate NEAR AI account is unnecessary for the OpenRouter route.

The direct API is NEAR’s confidential route

NEAR says requests made through OpenRouter pass through OpenRouter and a gateway outside NEAR’s confidential environment. Its announcement says the model runs on trusted-execution hardware, but that gateway lacks the attestation needed to make a claim about the complete request path. Developers who need NEAR’s confidential inference should call its direct API with a NEAR AI key.

OpenRouter’s provider comparison labels NEAR AI as zero retention and says it does not train on requests. Those labels address data handling by the inference provider. OpenRouter’s explanation of zero data retention says a request still travels through the router and the provider, while metadata and other services have their own controls. That is a different guarantee from a verified confidential path through every part of the request.

On the direct route, NEAR describes Intel TDX virtual machines paired with NVIDIA GPUs in confidential-computing mode, with an attestation report developers can check. Its direct catalogue also includes other models and services, including embeddings, reranking, image generation and speech to text; those products are separate from the single OpenRouter listing.

OpenRouter charges for credits, then for usage

OpenRouter passes through the selected provider’s inference price and charges its platform fee when credits are purchased. Its spend guide lists a 5.5 per cent fee on standard card top-ups, with an $0.80 minimum. A developer using the NEAR AI route pays from an OpenRouter balance; a developer using NEAR’s confidential direct route uses a NEAR AI account and its billing.

The addition gives existing OpenRouter users a way to choose NEAR AI for GLM 5.3 Flash without changing their application to a new provider API. The routing flag decides whether another provider may take over. The API path decides which privacy and verification guarantees apply.