GLM-5.3-Flash

Requesty's August 28, 2026 industry roundup documents a nine-day wave in which five labs shipped open-weight models with 1M token context windows, headlined by GLM-5.3-Flash from Z.ai on August 26. The post specifies 320B total parameters with 18B active, native multimodality, MIT licensing, training on Chinese AI chip The post's pricing ($0.15/$0.50) differs from Requesty's own managed model page ($0.07/$0.25), suggesting the blog figures reflect an earlier or upstream listing rather than current router rates. Social-listening mention-count metrics (506 to 978 week-over-week) are drawn from Requesty's internal corpus and are not ind

On August 20, 2026, an anonymous model called "Ox Alpha" appeared on OpenRouter and OpenCode with a 1M-token context window and text, image, and video input. Within days it became the most-used model on OpenRouter, processing roughly 23 trillion tokens, the largest launch in the platform's history at about 2.3 times th On August 26, 2026, Bloomberg confirmed that Ox Alpha was GLM-5.3-Flash from Zhipu AI (operating internationally as Z.ai), a 320B-total / 18B-active model and the first natively multimodal release in the GLM-5 series. MIT-licensed weights shipped to Hugging Face the same evening, and Zhipu's Hong Kong shares closed mor

AI Profit Boardroom confirms the OpenRouter model listing details for GLM-5.3-Flash: a mixture-of-experts model with 320 billion total parameters and 18 billion active per token, built on a hybrid sparse and linear attention architecture, with a listed context window of 1,310,720 tokens and up to 131,072 completion tok The piece documents Z.ai's launch-discount pricing, positioning GLM-5.3-Flash at a tenth of the flagship GLM-5.3 API rate, or a twentieth during a limited-time discount window, and traces the Ox Alpha anonymous-preview arc back to August 20, 2026. While the article wraps this technical content in promotional SEO framin

A Medium comparison piece confirms that Z.ai released GLM 5.3 Flash on August 26, 2026, twelve days after the flagship GLM 5.3 launch on August 14, 2026. Flash is described as a 320-billion-parameter mixture-of-experts model with 18 billion active parameters per token, open MIT weights, and native image and video input The piece reports that Z.ai built GLM 5.3 by scaling post-training on the GLM 5.2 base without changing the underlying base model, with every capability gain coming from additional training environments, task diversity, and compute, resulting in a 753-billion-parameter model. In contrast, Flash starts from a new base a

The Requesty provider page for sference's EU deployment of GLM-5.3-Flash documents the managed id sference/glm-5.3-flash with full 1M input context and 262K output, input priced at $0.20 per 1M tokens (cache read $0.07) and output at $0.50 per 1M, yielding a measured blended cost of $0.07 per 1M tokens including cache. The page positions sference as the EU-region managed option with full native multimodal capability and the full 1M context, contrasting with provider-specific deployment caveats seen elsewhere in the corpus. Workload cost examples (100K+10K at $0.025, 1M+100K at $0.25, 10M+1M at $2.50 before caching) are given to illus

Z.ai released GLM-5.3-Flash on August 26, 2026, as a 320B-total / 18B-active mixture-of-experts open-weight model under an MIT license, supporting multimodal inputs and a 1M-token context window. The model is positioned around delivering frontier-level capabilities at far lower inference cost, combining text, image, an Before the official reveal, GLM-5.3-Flash circulated as the anonymous "Ox Alpha" preview on platforms including OpenRouter and OpenCode, where Z.ai collected real-world feedback ahead of public release. Z.ai later confirmed Ox Alpha was the GLM-5.3-Flash preview, tying the anonymous traffic spike to the same weights th

GLM 5.3 Flash from Z.ai is a 320B-total / 18B-active MoE with a 1M-token context and native image input, released under an MIT license, while the GLM 5.3 flagship uses a 753B-total / 40B-active base without native multimodality. In independent Artificial Analysis testing, GLM 5.3 scores 60 on the Intelligence Index ver The comparison shows Flash winning on API price, cost-performance, image understanding, visual coding, browser agents, computer-use agents, and Coding Plan quota, while GLM 5.3 wins on overall intelligence, complex and terminal coding, long-horizon engineering tasks, output speed, and time-to-first-token parity. The ar

DataCamp profiles GLM-5.3-Flash as Z.ai's cost-optimized sibling to the GLM-5.3 flagship, codename "Ox Alpha," using a 320B-A18B MoE design with a 1M-token context window and native multimodal input. It scores 84.3 on Terminal-Bench 2.1, within striking distance of Claude Opus 4.8 (85.0) and behind GPT-5.6 Terra (87.4) Pricing sits at roughly $0.10 per 1M tokens blended, about one-ninth of GLM-5.3's $0.90 blended price, with trade-offs of around 49 tokens/sec output speed and a 306 GiB FP8 checkpoint that rules out lightweight local hosting. Qwen3.8-Flash-Next launched the same day, sharpening competition in the budget-tier segment.