GLM 5.3 Flash

GLM 5.3 Flash, released by Z.ai on August 26, 2026, is a 320-billion-parameter mixture-of-experts model that activates only 18 billion parameters per token, the first natively multimodal entry in the GLM-5 series and the first built on a new base rather than a post-training refresh of GLM 5.2. The article documents tha The write-up also reveals that GLM 5.3 Flash is "Ox Alpha," the anonymous model that had been quietly serving traffic on OpenRouter with a 1M-token context before Z.ai claimed it. The piece notes that despite the "Flash" name this is not a distilled variant but a distinct new-base MoE, and contrasts the launch against

This Yotta Labs comparison positions GLM 5.3 (August 14, 2026, API-only flagship, GLM 5.2 base with post-training) against GLM 5.3 Flash (August 26, 2026, 320B/18B MoE, MIT weights on Hugging Face, native image and video input, $0.15 in / $0.50 out, self-hostable on an 8-GPU node floor). The piece highlights that the t On the only shared test, DeepSWE v1.1, GLM 5.3 Flash reaches 63.4 versus the flagship's 66.9 (and GLM 5.2's 46.2), landing Flash within roughly 3.5 points of the flagship at roughly a tenth of the price. The article notes Flash spent a week running in stealth as "Ox Alpha" before Z.ai disclosed it, so it shipped with r

Released by Z.ai on August 26, 2026, GLM-5.3-Flash is positioned within the GLM-5 family as a frontier-capability model optimized for inference cost, with 320 billion total parameters and approximately 18 billion active parameters per token. The article confirms native multimodal support, a 1-million-token context wind The Medium write-up frames the release as part of an industry shift away from raw parameter scale toward intelligence-per-compute, arguing that GLM-5.3-Flash's 320B/18B MoE design delivers competitive capability at a fraction of the active compute. It re-states the Z.ai launch facts including the August 26, 2026 releas

The Ampere.sh head-to-head comparison reports that in independent Artificial Analysis testing GLM 5.3 scores 60 on the Intelligence Index versus 57 for GLM 5.3 Flash, while Flash activates only 18B parameters against the flagship's 40B and costs approximately one-ninth as much at normal API pricing ($0.15/M input, $0.5 The piece attributes category winners by use case, with GLM 5.3 winning overall intelligence, complex coding, long-horizon engineering, terminal coding, and output speed, while GLM 5.3 Flash wins on API price, cost-performance, image understanding, visual coding, browser agents, computer-use agents, and Coding Plan quo

The llm-stats tracker ranks GLM-5.3-Flash at number 17 on its composite LLM Stats Score, with a blended price of approximately $0.17 per million tokens and an LLM Stats Score around 50. Capability sub-rankings place the model in the Top 10% for Tool Calling (7 of 193), while landing in the middle tier for Reasoning (16 Individual benchmark entries recorded on the scorecard include GDPval-AA v2 at 1773/3000 (rank 2, evaluated by Artificial Analysis), CharXiv-R at 0.89/1 (rank 8, methodology by z.ai), and Terminal-Bench 2.1 results, with scores sourced from the model's scorecard, paper, or official blog posts. A Quality Tracker shows +

Artificial Analysis publishes an independent profile of GLM-5.3-Flash, confirming it as a 320B-total / 18B-active MoE model from Z.ai with an MIT license and weights hosted on Hugging Face. The page documents a 1M-token context window, native text and image input with text output, reasoning enabled, and pricing of $0.1 The profile adds comparative context for the GLM family: Flash lands below the GLM 5.3 flagship (60 vs 57 on the Intelligence Index per the cited Artificial Analysis numbers) but at roughly one-ninth the per-token price and with multimodal capability the flagship lacks. Artificial Analysis flags that a non-reasoning va