Z.ai releases GLM-5.3-Flash, a multimodal MoE model with 1M context
ByPulseAugur Editorial·[19 sources]·
Z.ai has launched GLM-5.3-Flash, a natively multimodal mixture-of-experts model with 320 billion total parameters and 18 billion active parameters per token. This model boasts a 1 million token context window and supports image and video inputs, all released under an MIT license with weights available on Hugging Face. It reportedly outperforms its predecessor, GLM-5.2, on coding benchmarks and rivals Claude Opus 4.8 in coding tasks, while being significantly cheaper to operate. The model's efficiency is attributed to a hybrid attention architecture and an IndexPool mechanism for managing long contexts, and it was initially tested anonymously on Chinese AI chips.
AI
IMPACT
Sets a new benchmark for cost-effective, long-context multimodal models, potentially accelerating adoption in coding and complex document analysis.
RANK_REASON
Frontier-lab model release with system card.
<p>Z.ai has released GLM-5.3-Flash, the first natively multimodal model in the GLM-5 series — a 320B-total / 18B-active MoE with a 1,048,576-token context window, MIT-licensed weights on Hugging Face, and API pricing at $0.15/M input and $0.50/M output. It scores 84.3 on Terminal…
Open-weight model zai-org/GLM-5.3 is climbing Hugging Face with 94.4k downloads in 30 days. Real-world usage, not just buzz. What’s driving that traction? https:// olud.ai/#leaderboard # OpenSource # AI # HuggingFace
Z.ai 26. srpna 2026 oficiálně vypustil model GLM-5.3-Flash, jde o první nativně multimodální model v řadě GLM-5 a zároveň nejlevnější schopný kódovací model, jaký firma dosud vydala. Jde o model typu mixture-of-experts s 320 miliardami parametrů celkem, z nichž je aktivních 18 mi…
Z.ai introduces GLM-5.3-Flash, a natively multimodal model with 320B total and 18B active parameters. It outperforms GLM-5.2 in coding benchmarks using a hybrid attention architecture that cuts long-context costs. At $0.045/task, it offers frontier intelligence at 1/10th the cost…
🤗 zai-org/GLM-5.3 is climbing Hugging Face right now: 151k downloads in 30 days, 1.6k likes. Open-weight text generation model with real traction. See where it ranks. https:// olud.ai/#leaderboard # OpenSource # AI # HuggingFace
<h1> GLM-5.3 Flash (Ox Alpha) Is Open Source: The $0.075/M Token Model That Hit #1 on OpenRouter in 6 Days — Full Price War Map </h1> <p><strong>An anonymous model appeared on OpenRouter, topped weekly usage at 11.6 trillion tokens, and turned out to be Z.ai's newest open-weight …
🤗 zai-org/GLM-5.3, an open-weight text generation model, is climbing Hugging Face with 94.4k downloads in 30 days and 1.5k likes. See what’s driving its real-world usage. https:// olud.ai/#leaderboard # OpenSource # AI # HuggingFace
GLM-5.3-Flash from zai-org has 379.3k downloads in 30 days, fewer stars per download than most open-weight models—meaning it gets used heavily without hype. https:// olud.ai/#leaderboard # OpenSource # AI # HuggingFace
GLM 5.3 Flash (Z.AI) is open-weight, runs 1.3M tokens of context, and prices at $0.08 in / $0.25 out per 1M. Few models pair that context with that cost — see the hourly data now. https:// olud.ai/latest.html # AI # LLM # OpenSource
An open-weight AI model just beat GPT-5.6 Luna on a major benchmark. 🤯 GLM-5.3-Flash: 57 GPT-5.6 Luna: 52 GLM-5.3-Flash also offers: 🔓 Open weights 📚 1M-token context 💰 Lower cost One benchmark doesn't mean it's better at everything — but the open AI race is getting seriously com…
Китайці випустили монстра GLM-5.3 Flash — аналог DeepSeek, який швидко набрав популярність на OpenCode та OpenRouter У моделі 320 млрд параметрів, вона випереджає популярні моделі за більшістю бенчмарків і водночас коштує у 10 разів дешевше. У кодингу та агентських завданнях майж…
🆕 GLM 5.3 Flash (batch) (Z.AI) arrives with open weights and a 1M-token context—priced at $0.15 in / $0.5 out. That context length at that cost is rare among open models. See the full hourly tracker: https:// olud.ai/latest.html # AI # LLM # OpenSource
🧠 Z.ai ha presentato # GLM 5.3-Flash, un modello da 320 miliardi di parametri totali che ne attiva solo 18 miliardi per ogni inferenza. 👉 I dettagli: https:// lnkd.in/p/eeZDdqxG ___ ✉️ 𝗦𝗲 𝘃𝘂𝗼𝗶 𝗿𝗶𝗺𝗮𝗻𝗲𝗿𝗲 𝗮𝗴𝗴𝗶𝗼𝗿𝗻𝗮𝘁𝗼/𝗮 𝘀𝘂 𝗾𝘂𝗲𝘀𝘁𝗲 𝘁𝗲𝗺𝗮𝘁𝗶𝗰𝗵𝗲, 𝗶𝘀𝗰𝗿𝗶𝘃𝗶𝘁𝗶 𝗮𝗹𝗹𝗮 𝗺𝗶𝗮 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿: https:// bit.…
Z.ai veröffentlicht GLM-5.3-Flash: 320B Parameter, Open Source und nur drei Punkte hinter dem größeren GLM-5.3 im Intelligence Index. Der gesamte Inference-Traffic lief ohne Nvidia-Hardware, was die Kosten auf ein Siebtel der Konkurrenz senkt. https:// the-decoder.de/chinesisches…
🔥 LM Studio se pone serio con la IA. Bionic ya integra GLM-5.3-Flash: multimodal, imágenes y hasta 1 millón de tokens de contexto, además de ser mucho más barato que GLM-5.2. ¿El próximo gran modelo para Mac? 👀 # AI # LMStudio
Pod koniec sierpnia Arabia Saudyjska zamieni się w technologiczne centrum świata. Piąta edycja konferencji LEAP to już nie tylko pokaz siły, ale realny punkt zwrotny dla globalnego AI i kapitału venture capital. # si # ai # sztucznainteligencja # wiadomości # informacje # technol…
Startup Z.ai zaprezentował GLM-5.3-Flash – natywnie multimodalny model o otwartych wagach, który oferuje wydajność klasy premium przy dziesięciokrotnie niższych kosztach. # si # ai # sztucznainteligencja # wiadomości # informacje # technologia https:// aisight.pl/technologia/gene…
Zhipu AI veröffentlicht GLM-5.3-Flash: 320B Parameter (18B aktiv) mit nativer Bild-Text-Modalität. Die Hybrid-Architektur kombiniert sparse und lineare Attention für effiziente Long-Context-Serving. MIT-Lizenz, FP8-Unterstützung und Deployment via SGLang/vLLM. https:// huggingfac…
