# GLM 5.3 Flash on Mistral: API Pricing, Context and Agent Support

Mistral has added Z.ai's existing model to its API, giving developers another low-cost option without changing their provider.

By Paras Madan | 2026-10-24


Mistral added Z.ai's GLM 5.3 Flash to its API on October 22, with input pricing of $0.15 per million tokens and support for agent workflows. The change is a new way to access an existing model, rather than a new GLM release.

For teams already using Mistral, that distinction is useful. They can evaluate another model inside an existing provider relationship instead of starting with a separate integration.

## GLM 5.3 Flash API pricing on Mistral

The [Mistral pricing page](https://docs.mistral.ai/inference/pricing) lists these standard USD rates per million tokens:

| Model | Input | Cached input | Output |
| --- | --- | --- | --- |
| GLM 5.3 Flash | $0.15 | $0.03 | $0.50 |
| Mistral Small 4 | $0.15 | $0.015 | $0.60 |
| GLM 5.3 | $1.40 | $0.14 | $4.40 |

Flash and Small 4 therefore have the same uncached input price, but different tradeoffs. Flash costs less for generated tokens, while Small 4 has a lower cached-input rate. A workload with substantial prompt reuse might favor a different model from one that produces long answers.

At the Flash rates, one million uncached input tokens plus one million output tokens costs $0.65 in model token charges. That is an illustration, not an estimate of a complete agent run, which can involve repeated calls and other services.

## API model ID and supported features

Use `zai-glm-5-3-flash` as the model identifier. Mistral's [model card](https://docs.mistral.ai/models/zai-glm-5-3-flash) marks it generally available, with a 1M-token context window and a maximum output of 128k tokens.

The card lists chat completions, structured outputs, function calling, document question answering, batching, Agents and Conversations, and built-in tools. It describes a multimodal mixture-of-experts model with 320 billion total parameters and 18 billion active parameters, served without Mistral modifications.

This makes it a candidate for a controlled comparison inside an existing agent application. Keep the task, tools and success criteria fixed, then compare completion quality, retries, latency and total token use rather than only the price of one response.

## What this availability change does not establish

The [dated changelog](https://docs.mistral.ai/resources/changelogs) confirms the October 22 addition. It does not establish that Mistral's deployment is faster or more reliable than other providers, or that Flash can replace a larger model on every workflow.

Z.ai publishes the model under an [MIT license](https://huggingface.co/zai-org/GLM-5.3-Flash/blob/main/LICENSE). That is separate from the cost and terms of using a hosted API. Nor should a third-party listing be read as a promise about a particular deployment region or data-handling arrangement.

The useful question is whether Flash can finish the same work with fewer dollars and no unacceptable loss in reliability. Provider convenience makes that experiment easier; it does not decide the result.

I'm Paras. I work at the intersection of AI, engineering and distribution.

I help technical founders turn product updates and engineering decisions into clear stories developers can use and share.

[Book a 30-minute call](https://cal.com/paras-madan/30min).