We use cookies. Cookies and browser storage keep you signed in and remember your preferences. With your permission we also measure how the site is used. You can any time. See our cookie policy.
z-ai/glm-5.3-flashx-legalRuns on your balance at the prices above. New accounts start with free usage.
Settings
Z.ai: GLM 5.3 FlashX Legal is a text model from Z.ai, built as the legal-domain variant of GLM-5.3-FlashX. It is a native multimodal model, meaning it was designed from the ground up to handle more than one input type rather than having vision bolted on afterwards. Z.ai adds legal-domain mid-training and post-training with reinforcement learning on top of the base FlashX model, which is what distinguishes this variant from the general-purpose release. If you are building anything that touches contracts, filings, correspondence or legal research, this is the version of the model Z.ai intends you to reach for.
The model accepts text, image and video as input, and returns text. That combination matters more than it might first appear: a scanned contract arrives as an image, a recorded deposition or hearing arrives as video, and a clause library arrives as text, and all three can go into the same model. The context window is 1,048,576 tokens, with a maximum output of 131,072 tokens. That input ceiling is large enough to hold entire document sets rather than single files, which changes how you structure a request. You can pass a full agreement, its amendments and the related correspondence in one call instead of chunking and stitching results together.
Practical uses follow directly from those capabilities. Reviewing long agreements against a playbook, extracting defined terms, obligations and dates from scanned or photographed documents, summarising recorded proceedings, comparing a draft against a standard template, and answering questions grounded in a corpus of filings are all natural fits. The large output allowance supports work that produces substantial text rather than a short answer, such as drafting a memo or a redline rationale. Because images and video are accepted alongside text, you can point the model at source material in whatever form your team actually stores it.
You call it through api14's OpenAI-compatible API using the model id "z-ai/glm-5.3-flashx-legal". Point your existing OpenAI client at the api14 base URL and pass that id in the model field. The request parameters it supports are include_reasoning, max_tokens, reasoning, reasoning_effort, response_format, structured_outputs, temperature, tool_choice, tools, top_k and top_p. The reasoning controls let you decide how much internal deliberation the model does before answering, which is useful when accuracy matters more than latency. Structured outputs and response_format are worth using when you need parsed fields rather than prose, and tools and tool_choice let the model call functions you define, so you can wire it into retrieval or document management systems.
One thing to weigh is that this is a legal-domain variant. If your workload is general text generation, summarisation or chat with no legal content, the general GLM-5.3-FlashX model is the more obvious choice, and the domain tuning here is unlikely to help. Output is text only, so anything requiring generated images or audio needs a different model. The supported parameter list does not include every option some OpenAI-compatible clients send, so check that your request only uses the parameters above before you debug a failure.
This community has no topics yet. Open the first one about Z.ai: GLM 5.3 FlashX Legal.
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api14.com/api/v1",
apiKey: process.env.API14_API_KEY,
});
const result = await client.chat.completions.create({
"model": "z-ai/glm-5.3-flashx-legal",
"messages": [
{
"role": "user",
"content": "Explain vector databases in two sentences."
}
]
});
console.log(result.choices[0].message.content);
include_reasoningmax_tokensreasoningreasoning_effortresponse_formatstructured_outputstemperaturetool_choicetoolstop_ktop_p