Z.ai offers the GLM model family through a chat app and a developer API billed per million tokens. GLM-5.3-Flash is listed at $0.15 input and $0.50 output, and the API is compatible with the OpenAI SDK.
Z.ai is the company behind the GLM family of language models. Its main site, z.ai, opens the Z.ai chat app, and its developer documentation describes an open platform where you register, top up a balance, create an API key and call models over a REST API. Official Python and Java SDKs are offered, and the API can also be reached through the OpenAI SDK, which makes it easier to switch existing code. The documentation lists GLM-5.3 as the latest flagship model, with improvements in software engineering and agent tasks, alongside GLM-5.3-Flash and FlashX, a multimodal coding model for visual programming, and capabilities such as thinking mode, function calling, tool streaming, structured output and context caching.
API pricing is per million tokens in US dollars. GLM-5.3-Flash costs $0.15 for input, $0.03 for cached input and $0.50 for output, GLM-5.3-FlashX costs $0.37 and $1.25, and the full GLM-5.3 and GLM-5.2 models cost $1.40 for input and $4.40 for output. Storage of cached input is listed as free for a limited time. Built-in web search costs $0.01 per use, image generation with GLM-Image is $0.015 per image, and agent products such as a slide and poster agent and a translation agent have their own rates.
Z.ai also sells a separate GLM Coding Plan subscription for use inside coding agent tools. No free API tier is stated on the pricing page, so you pay from the first token after topping up.
GLM-5.3-Flash is $0.15 per million input tokens and $0.50 per million output tokens, with cached input at $0.03. The full GLM-5.3 model is $1.40 input and $4.40 output per million tokens.
Yes. Z.ai's quick-start guide lists OpenAI SDK compatibility alongside its own Python and Java SDKs, so existing OpenAI-style code can point at the Z.ai endpoint.
Website: https://z.ai/
Company: Z.ai
Developer API for Z.ai's GLM models with OpenAI SDK compatibility and low Flash-tier rates; no free tier stated. GLM-5.3-Flash $0.15 in / $0.50 out per million tokens.
| Pricing model | Usage-based |
|---|---|
| Free tier | No free plan |
| Starting price | Not verified |
| Data residency | Not verified |
| Compliance (as stated by vendor) | Not verified |
| Integrations | Not verified |
| Public API | Yes |
| Self-hostable | Not verified |
| Official site | Not verified |
| Can a general AI assistant replace it? | For chatting, chat.z.ai is enough. Use the API when you want GLM models in your own code or coding agents and need to control per-token cost. |
| Last verified | 2026-10-12 by editor |
Open Interpreter is an open-source terminal agent, forked from Codex, that reads and edits files and runs commands inside an OS-level sandbox with approvals. It is free under Apache 2.0 and works with local models through Ollama or LM Studio.
Together AI is a cloud platform for running open models through a per-token serverless API, plus dedicated endpoints, GPU clusters and fine-tuning. Hosted GLM-5.3-Flash, for example, is listed at $0.15 input and $0.50 output per million tokens.
Kimi is Moonshot AI's assistant and model family, offered as the kimi.com agent workspace and as a pay-as-you-go API. The Kimi K2.7 Code model is listed at $0.95 input and $4.00 output per million tokens.
The Gemini API and Google AI Studio let developers build with Google's Gemini models for text, code, images, video and speech. A free tier covers some models; paid use is billed per million tokens, such as $0.30 in and $2.50 out on Gemini 3.5 Flash-Lite.
The Claude API is Anthropic's developer platform for building apps and agents on Claude models, billed per million tokens. Rates run from Claude Haiku 5.5 at $0.10 input and $0.50 output up to Claude Fable 5.1 at $10 and $50.
Google Antigravity is Google's agent-first development platform with a desktop app, IDE, CLI and SDK. The Individuals plan is free with weekly rate limits; higher limits come with Google AI Pro or Ultra.
Need the outcome rather than the software?