Every question we are asked about Ox Alpha and GLM-5.3-Flash — the same model, before and after its August 26, 2026 reveal — answered briefly, with a link to the full page. Facts come from the live OpenRouter listing, Z.ai’s announcement, and our verification rules.
Is Ox Alpha the same as GLM-5.3-Flash?
Yes. Ox Alpha was the codename Z.ai used while testing the model anonymously on OpenRouter between August 20 and August 26, 2026. On the 26th it was revealed and relisted under its real name, GLM-5.3-Flash. Same model, same specifications, new name and a price. More →
Who made Ox Alpha?
Z.ai, also known as Zhipu AI, confirmed on August 26, 2026. Independent testers had identified the GLM family days earlier through tokenizer and error-code fingerprinting. More →
What is GLM-5.3-Flash?
A native multimodal reasoning model from Z.ai with 320 billion total parameters and 18 billion active per token, built for coding and long-horizon agent work. It has a 1,048,576-token context window, accepts text, image and video input, and uses a hybrid sparse and linear attention architecture. More →
Is Ox Alpha still free?
No. The free preview ended on August 26, 2026. GLM-5.3-Flash is now paid, at promotional rates until September 9, 2026. More →
How much does GLM-5.3-Flash cost?
$0.075 per million input tokens, $0.015 per million cached input tokens, and $0.25 per million output tokens — promotional pricing until 24:00 UTC+8 on September 9, 2026. After that the list price applies: $0.15, $0.03 and $0.50 respectively. More →
Is GLM-5.3-Flash open source?
Yes. Z.ai released the weights on Hugging Face as zai-org/GLM-5.3-Flash under the MIT licence, which permits commercial use, modification and redistribution. Documented deployment paths include SGLang, vLLM, TokenSpeed, KTransformers and others. More →
Does stealth/ox-alpha still work?
No. The ID was delisted at the reveal and no redirect exists, so calls to it fail rather than rerouting. Replace it with z-ai/glm-5.3-flash. More →
How do I use GLM-5.3-Flash?
Create an OpenRouter key, add credit, then either paste the key into the browser chat on this site or call the OpenRouter API with model id z-ai/glm-5.3-flash. You can also call Z.ai directly with the id glm-5.3-flash. More →
Can I still use it for free?
Not through OpenRouter’s paid listing. Two routes remain: run the MIT-licensed weights on your own hardware, or use one of OpenRouter’s rotating free models, which the chat on this site offers in its model switcher without any credit on your account. More →
Do I need an account to chat on this site?
No account here. You paste your own OpenRouter key once; it is stored only in your browser and requests go directly from your browser to OpenRouter. This site never sees your key or your prompts. More →
What is the GLM-5.3-Flash context window?
1,048,576 tokens of input, with up to 131,072 tokens of output per response. Both figures are unchanged from the stealth listing. More →
Does it support images and video?
Yes. Text, image and video input are accepted; output is text only. Audio input is rejected. Multimodality is native to the base model rather than added through a separate vision encoder. More →
Does it support tool calling and JSON output?
Yes. The listing declares tools, tool_choice and response_format, plus reasoning, reasoning_effort, include_reasoning, temperature, top_p, top_k, seed, logprobs and max_tokens. More →
Can I turn off reasoning?
No. Reasoning is mandatory and enabled by default at max effort. You can lower it to high or low, which noticeably reduces the pause before the first token. More →
What are its benchmark scores?
Z.ai reports 63.4 on DeepSWE v1.1, 48.8 on AutomationBench and 84.3 on Terminal Bench 2.1 — vendor claims, not independently reproduced. During the stealth preview an independent tester measured roughly 63% on the full 113-task DeepSWE set, which corroborates the DeepSWE figure closely. An earlier viral “80%” came from a 10-task subset and was superseded. More →
Is it better than ChatGPT, Claude, Gemini or Grok?
Not on the hardest reasoning tasks, where the leading commercial flagships remain ahead. It is extremely hard to beat on cost per token, it matches or exceeds most of them on context length, and it is the only one of the group with openly licensed weights. More →
Is it safe to send my data to it?
The Stealth Program terms no longer apply; usage falls under the standard OpenRouter and Z.ai terms, and the processor is identified rather than anonymous. Apply the same diligence you would to any third-party API. If your data cannot leave your infrastructure, the MIT-licensed weights make self-hosting a genuine option. More →
Can I use it in Cursor, Cline or Aider?
Yes — any tool that accepts an OpenAI-compatible base URL. Use https://openrouter.ai/api/v1, your OpenRouter key, and model id z-ai/glm-5.3-flash. If a tool is still configured with the old stealth ID it will now fail. More →
What is a stealth model?
An unreleased model that OpenRouter lists anonymously under a codename so its developer can gather real-world usage before launch. Seven codenames have now been revealed, Ox Alpha included; none has ever simply vanished unidentified. More →
Is this site affiliated with Z.ai or OpenRouter?
No. OxAlpha.chat is an independent, unofficial site. All trademarks belong to their owners. More →
Sources and verification
Last substantively reviewed: August 27, 2026. Reviewed by: OxAlpha.chat Editorial Team.
How we verified this page
Answers were checked against first-party model documentation, live platform metadata, official pricing and the published license. Time-sensitive answers are dated and should be rechecked at the linked source.
Primary sources
- Official model documentation: Z.ai GLM-5.3-Flash overview — supports the identity, context, multimodal and capability answers.
- Official pricing: Z.ai API pricing — supports the promotional and standard token rates.
- Platform metadata: OpenRouter GLM-5.3-Flash listing — supports the current slug, routes and live availability.
- Published checkpoint: Official Hugging Face model card — supports the open-weight and self-hosting answers.
- License: Official MIT license file — provides the exact reuse terms for the weights.
Found a factual error or a changed price? Send us the page URL and supporting source; corrections follow our editorial policy.