GLM-5.3-Flash

The excerpt reports that Z.ai released GLM 5.3 Flash on August 26, 2026, with MIT-licensed weights on Hugging Face on day one. The model is a 320-billion-parameter mixture of experts activating 18 billion parameters per token, the first natively multimodal model in the GLM-5 series, and the first built on a new base si The excerpt confirms that GLM 5.3 Flash is the same model previously previewed anonymously as "Ox Alpha" on OpenRouter and OpenCode, and that Z.ai's own announcement framed it as a 320B-A18B model released under the MIT License that ran entirely on Chinese AI chips. The article highlights that despite GLM 5.3's flagshi

According to the excerpt, on August 20, 2026, an anonymous listing called "Ox Alpha" appeared on OpenRouter and OpenCode with no owner listed, a 1M-token context window, and text, image, and video input. Six days later it had become the most-used model on OpenRouter, processing roughly 23 trillion tokens in what the ex The excerpt states that on August 26, 2026, Bloomberg confirmed Ox Alpha as GLM-5.3-Flash from Zhipu AI (operating internationally as Z.ai), with 320 billion total parameters, 18 billion active per token, making it the first natively multimodal model in the GLM-5 series and the cheapest model at its score on Artificial

The South China Morning Post reports that Zhipu AI's Hong Kong-listed shares jumped after the viral Ox Alpha model was revealed as GLM-5.3-Flash, with the company stating the system ran entirely on a cluster of 100,000 domestically produced chips during a high-profile stealth trial. The launch is framed as evidence tha The article, published August 27, 2026, positions GLM-5.3-Flash as Zhipu AI's latest open-weight model that was previously code-named Ox Alpha, connecting the model's anonymous OpenRouter run to its formal identity and to broader China-tech sector dynamics. The coverage emphasizes the strategic significance of running

Z.ai released GLM-5.3-Flash on August 26, 2026, as the first natively multimodal model in the GLM-5 series, described by Z.ai as the cheapest capable coding model the lab has shipped. The model uses a mixture-of-experts architecture with 320B total parameters and 18B active per token, supports a 1,048,576-token context According to Z.ai reports cited in the article, GLM-5.3-Flash beats GLM-5.2 across benchmarks and real workloads at roughly one-tenth the price, while landing within half a point of Claude Opus 4.8 on Z.ai's internal coding benchmark. The default FP8 checkpoint is roughly 306 GiB of weights before KV cache, and the cur

Local AI Zone's technical deep-dive describes Z.ai's quiet August 26, 2026 ship of GLM-5.3-Flash as a 320B-parameter mixture-of-experts model activating only 18B per token, running natively in FP8 with a 1,048,576-token context window. The post notes that, per the model card, the model "starts from a newly trained base For twelve days before launch the model ran anonymously on OpenRouter as "Ox Alpha," beating GLM-5.2 across six coding and agentic benchmarks at roughly one-tenth the price and approaching Claude Opus 4.8 on long-horizon agent tasks. Weights ship under MIT on Hugging Face, the API is priced at $0.15/$0.50 per million t

DataCamp reports that Z.ai released GLM-5.3-Flash as a cost-optimized sibling to the GLM-5.3 flagship, revealed publicly after an anonymous "Ox Alpha" preview period. The model uses a 320B-A18B mixture-of-experts configuration, supports a 1M token context window, and is natively multimodal — the first such model in the GLM-5.3-Flash 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), with Artificial Analysis rating it at an intelligence index of 57 versus 60 for GLM-5.3. Blended price is roughly $0.10 per 1M tokens, about a tenth of GLM-5.3's $0.90, with trade-offs of