moe — AI Digest

30 апр DeepSeek V4: official open-source release with Day-0 adaptation for Huawei Ascend DeepSeek models-llm
7 июл Tencent Officially Releases Hunyuan Hy3: 295B MoE Model with Agent and Reasoning Capabilities Tencent models-llm
28 июл Moonshot Kimi K3 Open-Weight Release: 2.8T Parameter MoE Model on Hugging Face Moonshot AI models-llm
2 июн MiniMax Releases M3: Open-Weight Frontier Model with 1M-Token Context and MSA Architecture MiniMax models-llm
8 июн NVIDIA Nemotron 3 Ultra: Open 550B MoE Model Now Available for Agentic Workloads NVIDIA models-llm
10 июн MiniMax M3 Open Weights Released: 1M Context, MoE, Frontier Coding MiniMax models-llm
17 июн Zhipu AI Open-Sources GLM-5.2 Under MIT License with 1M Token Context Zhipu AI models-llm
19 июн Zhipu AI Releases GLM-5.2 Open Weights: 753B MoE with 1M-Token Context under MIT License Zhipu AI / Z.ai models-llm
26 июн ByteDance Launches Doubao-Seed-2.1 Pro Flagship LLM at FORCE Conference ByteDance / Doubao models-llm
7 июл Sber Releases GigaChat 3.5 Ultra: 432B Open-Source MoE Flagship Sber models-llm
9 июл Tencent Releases Hunyuan Hy3: 295B Open-Weight MoE Model Under Apache 2.0 Tencent models-llm
16 июл Thinking Machines Lab Releases Inkling: 975B Open-Weight Multimodal MoE Thinking Machines Lab models-llm
17 июл Moonshot AI launches Kimi K3, a 2.8T-parameter open-weight model Moonshot AI models-llm
29 авг Zhipu open-weights GLM-5.3 (753B MoE), claiming the open-weights coding frontier Zhipu models-llm
30 авг Tencent open-sources Hy4 preview, a 770B MoE flagship with 1M context Tencent Hunyuan models-llm
11 сен DeepSeek V4.1 Flash goes GA with open MIT weights and V4 Pro retirement DeepSeek models-llm
28 авг Tencent open-sources Hy4 preview: 770B / 49B-active MoE with Gated DSA + iHC, 1M context, Apache 2.0 Tencent models-llm
9 мая Zyphra Releases ZAYA1-8B: Open Reasoning MoE Model Trained on AMD Hardware Zyphra models-llm
20 мая Lance: 3B Unified Multimodal Model for Understanding, Generation, and Editing (314 HF upvotes) ByteDance Research research
4 июн JetBrains Open-Sources Mellum2: 12B MoE Coding Model for Multi-Model Pipelines JetBrains models-llm
10 июн Cohere North Mini Code: 30B Apache-2.0 MoE Coding Model for Agentic Workflows Cohere models-llm
16 июн Kimi K2.7-Code HighSpeed: 6× Throughput for Production Coding Agent Pipelines Moonshot AI models-llm
24 июн Prime Intellect Releases prime-rl v0.6.0 for Agentic RL on Trillion-Parameter MoE Models Prime Intellect research
26 июн DeepReinforce Releases Ornith-1.0: Open-Source Coding Models That Learn Their Own RL Scaffolds DeepReinforce tools
30 июн DeepSeek V4 Stable Release Set for Mid-July with First Time-of-Day API Pricing DeepSeek models-llm
3 июл DeepSeek Confirms V4 Official Launch for Mid-July with Peak-Time API Pricing DeepSeek models-llm
9 июл NVIDIA Releases Nemotron-Labs-Audex-30B-A3B: Unified Audio-Text MoE Model NVIDIA audio
16 июл DeepSeek V4 Graduates from Preview to General Availability with Peak-Hour API Pricing DeepSeek models-llm
20 июл Alibaba previews Qwen3.8-Max, a 2.4-trillion-parameter multimodal flagship Alibaba / Qwen models-llm
13 авг DeepSeek V4 Pro exits preview as V4-Pro-0813 with major coding benchmark gains DeepSeek models-llm
6 сен Saudi HUMAIN unveils humain-m3, a 428B Arabic frontier model built by MiniMax MiniMax models-llm
18 сен PrismML ships Bonsai 2 27B: ternary weights near-lossless to the teacher PrismML models-llm
11 июн Kwai Keye-VL-2.0: Open-Source 30B MoE Multimodal Model with 256K Context for Long Video Kwai research
14 июн Moonshot AI Releases Kimi K2.7-Code: 1T-Parameter Open-Weight Coding Model with Vision Moonshot AI models-llm
16 июл Ring-Zero: Scaling Zero RL to 1 Trillion Parameters with Emergent Reasoning Behaviors Ant Group research
28 июл SLAI T-Rex: Full-Parameter Post-training of the DeepSeek-V4 Family on Ascend SuperPOD SLAI research
29 июл Kimi K3 Technical Report: Kimi Delta Attention and Stable LatentMoE Architecture Detailed Moonshot AI research
5 авг SwanTale: Unified Multi-Speaker Speech and Audio Generation for Instruct and Zero-Shot Tasks ByteDance research
1 сен DeepSeek open-sources V4-Flash-Vision-Exp, its first multimodal V4 model, under MIT DeepSeek models-llm
14 июн vLLM Adds Day-0 Support for MiniMax M3 Open Weights with 1M-Context Sparse Attention MiniMax tools
8 июл Kimi K2.7 Code (1T params, 32B active) expands to GitHub Copilot Business and Enterprise Moonshot AI tools
23 июл SLAI T-Rex: Full-Parameter Post-training of the DeepSeek-V4 Family on Ascend SuperPOD SLAI research
27 авг Zhipu releases GLM-5.3-Flash, first natively multimodal GLM-5 model Zhipu AI / Z.ai models-llm
27 авг Qwen3.8-Flash-Next released as experimental Qwen4 architecture preview Alibaba (Qwen Team) models-llm
1 сен Qwen team details the Qwen3.8-Next architecture: hybrid attention, n-gram embeddings, Muon Qwen (Alibaba) research
6 сен Yandex merges VLM and LLM into a single omnimodel powering Alice AI Yandex models-llm
12 сен Ant Group open-sources Ling-3.0-flash-Fin, its first finance-enhanced model Ant Group models-llm
18 сен DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression DeepSeek research
14 июн Zhipu AI Releases GLM-5.2: 744B MoE with 1M-Token Context and Coding-First Design Zhipu AI models-llm
26 июн GLM-5.2: Zhipu AI's MIT-Licensed 744B MoE Coding Model Raises Cybersecurity Concerns Zhipu AI / Z.ai models-llm
3 авг Alibaba Makes Qwen3.8-Max Widely Accessible via API, Launches QwenWork Public Beta Alibaba / Qwen models-llm
17 авг Intern-S2-Mobius: Foundation Model with Decoupled Knowledge and Reasoning Shanghai AI Lab research
2 сен SMELT: looped MoE transformers match baseline scaling at matched compute ByteDance Seed research
18 сен Infinite-Parameter LLMs: arXiv paper proposes generating weights from live data Independent researchers research
29 апр Sber unveils Kandinsky 6.0 Image — flagship image generation model Sber image
3 июл ELDR: Expert-Locality-Aware Routing Cuts MoE Serving Latency by up to 14% Microsoft Research research
14 июл llama.cpp Adds Tencent Hunyuan 3, Minimax2 Eagle3 Speculative Decoding, and SYCL Fused MoE tools
20 июл xHC: Expanded Hyper-Connections scale residual-stream parallelism past prior limits Shanghai Jiao Tong University / Xiaohongshu / USTC / Peking University / CUHK research
28 авг llama.cpp b10660 adds Qwen3.8-Flash-Next (qwen4exp) architecture: hyper-connections, gated delta net layers, MoE, PLE n-gram embedding llama.cpp tools