GLM-5.3-Flash

Linas Beliūnas's newsletter traces how an anonymous listing called "Ox Alpha" appeared on OpenRouter and OpenCode on August 20, 2026, with a 1M-token context window and free, no-owner-listed access. Within six days it became the most-used model on OpenRouter, processing roughly 23 trillion tokens (about 2.3x the next m The piece notes that GLM-5.3-Flash was described as the cheapest model at its score on Artificial Analysis's Intelligence Index, with MIT-licensed weights landing on Hugging Face the same evening as the reveal. Zhipu's Hong Kong shares closed more than 12% higher the following day, roughly 10x its January IPO price. Th

Z.ai released GLM-5.3-Flash on August 26, 2026, as a 320-billion-parameter mixture-of-experts model that activates 18 billion parameters per token, built on a newly trained base rather than a post-train of GLM 5.2's 744B foundation. It is the first natively multimodal model in the GLM-5 series, supporting a 1M-token co The article also reveals that GLM-5.3-Flash was the anonymous "Ox Alpha" model that had been quietly serving traffic on OpenRouter for roughly six days before launch, drawing attention for its 1M-token context window. Despite the launch, Z.ai's promised open-weight release of the flagship GLM 5.3's 744B parameters rema

AI Profit Boardroom's guide reiterates Z.ai's August 26, 2026 release of GLM 5.3 Flash as a 320B-parameter, 18B-active MoE with native multimodality (text, image, video input; text output) and a 1,310,720-token context window with up to 131,072 completion tokens. The model supports tool calling and JSON-formatted outpu The piece retells the Ox Alpha backstory: the anonymous model appeared on a third-party API platform on August 20 with a 1M-token context window and free access, was fingerprinted back to the GLM family within 48 hours, and was confirmed by Z.ai's August 26 announcement. The article focuses on access paths (open weight

Greek Ai's Medium piece confirms Z.ai's release of GLM-5.3-Flash on August 26, 2026, as a 320B-parameter MoE with 18B active parameters, multimodal inputs (text, image, video), a 1M-token context window, and MIT-licensed open weights. Z.ai positioned the model around the idea of "frontier-level capabilities without fro The article details the Ox Alpha provenance: before the official reveal, Z.ai had previewed GLM-5.3-Flash anonymously on OpenRouter and OpenCode to gather real-world developer feedback. While thinner on architectural depth than other candidates, the piece reiterates core specifications tied to Z.ai's announcement and i

Local AI Zone's deep dive confirms Z.ai's August 26, 2026 release of GLM-5.3-Flash as a 320B-total/18B-active MoE with hybrid sparse-and-linear attention, native FP8 operation, and a 1,048,576-token context window, marking it as the first natively multimodal model in the GLM-5 series. Weights shipped under MIT on Huggi The piece details benchmark results, noting that GLM-5.3-Flash beats GLM-5.2 across six coding and agentic benchmarks at roughly one-tenth the price and approaches Claude Opus 4.8 on long-horizon agent tasks. It also covers the Ox Alpha reveal: for about a week before launch, the anonymous model topped usage charts on