{"slug": "z-ai-releases-glm-5-3-flash-a-320b-a18b-natively-multimodal-moe-with-a-1m-token", "title": "Z.ai Releases GLM-5.3-Flash: A 320B-A18B Natively Multimodal MoE With a 1M-Token Context", "summary": "Z.ai released GLM-5.3-Flash, a 320B-total-parameter mixture-of-experts model with 18B active parameters, a 1,048,576-token context window, and native image and video input, under an MIT license with weights on Hugging Face. The model, which Z.ai reports beats GLM-5.2 at roughly one-tenth the price and lands within half a point of Claude Opus 4.8 on its internal coding benchmark, was served entirely on domestically produced Chinese AI chips during its first week as 'Ox Alpha' on OpenCode and OpenRouter.", "body_md": "Z.ai has released [GLM-5.3-Flash](https://z.ai/blog/glm-5.3-flash), the first natively multimodal model in the GLM-5 series and the cheapest capable coding model the lab has shipped. It is a mixture-of-experts model with **320B total parameters and 18B active per token**, a **1,048,576-token context window**, and image and video input — released under an **MIT license** with weights on [Hugging Face](https://huggingface.co/zai-org/GLM-5.3-Flash). According to Z.ai reports, it 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 its internal coding benchmark. The model spent its first week running anonymously as *“Ox Alpha”* on OpenCode and OpenRouter, served entirely on domestically produced Chinese AI chips.\n\n**Is it deployable?**\n\n**Yes**, **on two tracks.** The weights are live on [Hugging Face](https://huggingface.co/zai-org/GLM-5.3-Flash) under an **MIT license**, and a hosted API is already priced and serving.\n\n**Which companies can realistically self-host?** Not everyone. The default FP8 checkpoint is roughly[306 GiB of weights before KV cache](https://recipes.vllm.ai/zai-org/GLM-5.3-Flash), and the current vLLM path supports NVIDIA Hopper and newer only. That puts self-hosting in reach of mid-size and large orgs with at least an 8-GPU node (or a GB200 tray at TP4), plus AI-native startups renting GPU capacity. Everyone below that line consumes it as an API — where the economics, not the hardware, are the story.**Industries with immediate fit**: software and devtools, IT/BPO automation, financial services and insurance document operations, enterprise BI and back-office knowledge work, e-commerce and any team shipping UI at volume.**Applications:** repo-scale coding agents, terminal and browser/computer-use agents, million-token log and contract analysis, UI regression checking from screenshots, and spreadsheet/deck/dashboard reasoning that would otherwise need an OCR-to-text pipeline.\n\n**The architecture is where the efficiency comes from**\n\nGLM-5.3-Flash starts from a newly trained base model on a **30T-token multimodal corpus**. **Three changes are worth knowing:**\n\n**Hybrid attention**: For the first time in the GLM series, Z.ai combines linear and sparse attention. Per the[vLLM recipe](https://recipes.vllm.ai/zai-org/GLM-5.3-Flash), the 45-layer language model interleaves**KDA linear-attention layers with NoPE sparse MLA layers**, routes each token through** 8 of 288 experts**, and ships** native FP8 weights**plus one MTP draft layer. Linear attention handles local dependency; sparse attention retrieves the globally relevant context.**IndexPool**: At million-token context, retrieval itself becomes the bottleneck. IndexPool compresses groups of indexer key vectors through weighted pooling to hold down latency and memory. Z.ai reports**~3× less attention compute and a 4.4× smaller KV cache** versus GLM-5.3.**mHC**: The model adopts** Manifold-Constrained Hyper-Connections**to improve scaling efficiency. Against GLM-4.5, at similar total parameter count, GLM-5.3-Flash roughly halves both activated parameters and layer count.\n\n**Benchmarks**\n\nMost numbers mentioned in the table below are Z.ai-reported and the harnesses differ per test — the model card’s [footnotes](https://huggingface.co/zai-org/GLM-5.3-Flash) specify temperature, context limits and judge models per benchmark, so treat cross-model comparisons as setup-dependent.\n\n| Benchmark | GLM-5.3-Flash | Reference |\n|---|---|---|\n| Terminal-Bench 2.1 | 84.3 | Opus 4.8: 85.0 · GPT-5.6 Terra: 87.4 |\n| DeepSWE v1.1 | 63.4 | GLM-5.2: 46.2 |\n| AutomationBench | 48.8 | GLM-5.2: 26.2 |\n| HLE | 55.3 | — |\n| OfficeQA Pro | 62.4 | ahead of Opus 4.8 |\n| Z.ai Code Bench v1.0 (max) | 29.0 | Opus 4.8: 29.5 |\n\nIndependently, [Artificial Analysis](https://artificialanalysis.ai/models/glm-5-3-flash) scores it **57 on the Intelligence Index**, with **48.7 output tokens/sec** and **1.52s TTFT** on Z.ai’s API — strong intelligence-per-dollar, but slow and verbose. Vision is the weak flank: it trails Gemini 3.7 Flash on BabyVision and MVbench.\n\n**The serving story is the underreported part**\n\nZ.ai states the entire Ox Alpha preview ran on **domestically produced Chinese AI chips**, using a custom SGLang-based engine that disaggregates encoding, prefill and decoding, and reports a **3× end-to-end serving improvement** across tens of thousands of accelerators.\n\n**Pricing and access**\n\nStandard API pricing is **$0.15/M input, $0.03/M cached input, $0.50/M output**. Z.ai reports a score of 57 on Artificial Analysis Intelligence Index v4.1.1 at **$0.045 per task** on the discounted tier. The model is live for **all GLM Coding Plan tiers** — Lite ($18/mo), Pro ($80), Max ($168) — at\n\n**3× the usable quota of GLM-5.3**, and its multimodal capabilities surface in\n\n**ZCode** through Browser Use and Computer Use. Local serving is supported on\n\n**SGLang, vLLM, TokenSpeed and KTransformers**.\n\n**Key Takeaways**\n\n- 320B-A18B natively multimodal MoE, 1M context, MIT-licensed weights on Hugging Face.\n- Hybrid KDA linear + NoPE sparse MLA attention: ~3× less attention compute, 4.4× smaller KV cache.\n- 84.3 Terminal-Bench 2.1 and 63.4 DeepSWE v1.1 — near Opus 4.8, well past GLM-5.2.\n- $0.15/$0.50 per M tokens; 3× GLM-5.3 quota for every GLM Coding Plan tier.\n- Self-hosting needs ~306 GiB FP8 weights on Hopper-or-newer; everyone else uses the API.\n\nCheck out the ** Model Weights** and\n\n**Blog****.** Also, feel free to follow us on\n\n**and don’t forget to join our**[Twitter](https://x.com/intent/follow?screen_name=marktechpost)\n\n**and Subscribe to**\n\n[150k+ML SubReddit](https://www.reddit.com/r/machinelearningnews/)**. Wait! are you on telegram?**\n\n[our Newsletter](https://magic.beehiiv.com/v1/f5e63dd4-5653-4f09-83e2-321a8b1ba526?email={{email}})\n\n[now you can join us on telegram as well.](https://t.me/machinelearningresearchnews)Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? [Connect with us](https://forms.gle/wbash1wF6efRj8G58)\n\nAsif Razzaq is the CEO of Marktechpost Media Inc.. As a visionary entrepreneur and engineer, Asif is committed to harnessing the potential of Artificial Intelligence for social good. His most recent endeavor is the launch of an Artificial Intelligence Media Platform, Marktechpost, which stands out for its in-depth coverage of machine learning and deep learning news that is both technically sound and easily understandable by a wide audience. The platform boasts of over 2 million monthly views, illustrating its popularity among audiences.", "url": "https://wpnews.pro/news/z-ai-releases-glm-5-3-flash-a-320b-a18b-natively-multimodal-moe-with-a-1m-token", "canonical_source": "https://www.marktechpost.com/2026/08/26/z-ai-releases-glm-5-3-flash-a-320b-a18b-natively-multimodal-moe-with-a-1m-token-context/", "published_at": "2026-08-26 21:33:23+00:00", "updated_at": "2026-08-26 21:50:24.276529+00:00", "lang": "en", "topics": ["large-language-models", "generative-ai", "ai-research", "ai-products", "ai-infrastructure"], "entities": ["Z.ai", "GLM-5.3-Flash", "Hugging Face", "Claude Opus 4.8", "OpenCode", "OpenRouter", "vLLM", "Artificial Analysis"], "alternates": {"html": "https://wpnews.pro/news/z-ai-releases-glm-5-3-flash-a-320b-a18b-natively-multimodal-moe-with-a-1m-token", "markdown": "https://wpnews.pro/news/z-ai-releases-glm-5-3-flash-a-320b-a18b-natively-multimodal-moe-with-a-1m-token.md", "text": "https://wpnews.pro/news/z-ai-releases-glm-5-3-flash-a-320b-a18b-natively-multimodal-moe-with-a-1m-token.txt", "jsonld": "https://wpnews.pro/news/z-ai-releases-glm-5-3-flash-a-320b-a18b-natively-multimodal-moe-with-a-1m-token.jsonld"}}