GLM-5.3 Flash (Z AI)

GLM-5.3-Flash (API id glm-5.3-flash) is Z.ai's efficiency-tier release positioned next to GLM-5.3, with Artificial Analysis measuring it at 57 on its Intelligence Index at a blended $0.10 per 1M tokens ($0.09 per Index task) and ranking it 4 of 111 open-weights models on AA's intelligence-vs-cost Pareto frontier. It is Documented specs include a 1M-token context window with up to 128K output, a reasoning model with thinking always on that cannot be disabled on the direct API, list pricing of $0.15 in / $0.50 out per 1M tokens with cached input at $0.03, and a 50% launch promo ($0.075 / $0.25) running until September 9, 2026. Artifici

According to eesel AI's explainer, Z.ai (formerly Zhipu AI) shipped GLM-5.3-Flash on August 26, 2026 under the tagline "Frontier Intelligence, Flash Cost," as the first natively multimodal model in the GLM-5 series — image and video go in alongside text, text comes out. It is a mixture-of-experts model with 320B total The same article documents the "Ox Alpha" stealth launch: before release the model was tested anonymously on OpenCode and OpenRouter as "ox-alpha," later revealed to be GLM-5.3-Flash, where Z.ai claims it became the most popular model of that week. The piece notes pricing of $0.15/$0.50 per million input/output tokens

On August 27, 2026, Bitdeer AI Model Studio announced GLM-5.3-Flash is live on its serverless inference platform, giving enterprise users an alternative to OpenRouter's anonymous routing for calling the newly named Z.ai model. Bitdeer's post recounts the Ox Alpha stealth arc: Ox Alpha appeared on OpenRouter on August 2 Bitdeer describes the three architectural changes that explain why the stealth preview held up under production load: hybrid sparse + linear attention (the first in the GLM series, combining linear attention for local dependencies with sparse attention for global context retrieval, sharply reducing long-context serving

GLM 5.3 Flash was released by Z.ai on August 26, 2026 with MIT-licensed weights published on Hugging Face the same day, surprising the community that had been expecting the open-weight release of the 743B GLM 5.3 flagship. The model is a 320B total / 18B active mixture-of-experts built on a newly trained base rather th GLM 5.3 Flash is also the reveal of "Ox Alpha," the anonymous model that had been serving traffic on OpenRouter and drawing attention for its 1M-token context window; Ox Alpha hit #1 on OpenRouter and more than doubled DeepSeek usage before being named. However, GLM 5.3's own 744B weights, which Z.ai's Hugging Face org

According to AI/TLDR's model profile, Z.ai released GLM-5.3-Flash on August 26, 2026 as the Flash-tier entry in the GLM-5 series and its first model with vision built in natively rather than bolted on. It accepts text, images, video, and files while returning text, carries a 1M-token context window, and runs only in th The same article details the efficiency architecture: GLM-5.3-Flash pairs linear attention, which captures local dependencies through state modelling, with sparse attention that retrieves global context through a lightweight indexer. Z.ai reports this cuts attention compute by 3.01× and KV cache size by 4.44× versus GL

On August 26, 2026 (evening), Z.ai named its OpenRouter stealth model Ox Alpha as GLM-5.3-Flash — a 320B-A18B model under MIT license with a 1M multimodal context window and API pricing of $0.15/$0.50 per million tokens — confirming weeks of community speculation with serving-layer forensics showing a Java stack trace, The GLM-5.3 flagship launch on August 14, 2026 was framed around coding and cybersecurity rather than vision, with Artificial Analysis's independent evaluation later tying GLM-5.3 at 60 on its Intelligence Index alongside Kimi K3. However, Z.ai's Hugging Face placeholder for zai-org/GLM-5.3 listing August 28 as the rel