Parameter– tag –
-
Samsung Research Unveils AnySimLite: Sub-700KB, Sub-30ms On-Device AI Matching 7B-Parameter LLMs for Speech Classification
Samsung Research South Korea Overview Samsung Research has introduced AnySimLite, a lightweight few-shot similarity encoder capable of delivering state-of-the-art performance for multiple on-device speech-adjacent classification tasks. O... -
Open-Source LLMs GLM-5.2, Llama 4 Maverick, and Kimi K3 Lead Benchmarks in September 2026
Thunder Compute Global Overview In the September 2026 open-source LLM rankings, GLM-5.2, Llama 4 Maverick, and Kimi K3 demonstrated leading performance across various domains. The 744B MoE GLM-5.2 (40B parameters) excelled in reasoning b... -
Mistral AI Releases “Mistral-Medium-Next” Open-Weight Model Under Apache 2.0 License: 70 Billion Parameters, Optimized for Efficient Inference on Commodity Hardware with Competitive Performance
Mistral AI Blog France Overview Mistral AI has announced the immediate release of "Mistral-Medium-Next," its latest open-weight large language model, under an Apache 2.0 license. This 70 billion parameter model demonstrates competitive p... -
NVIDIA Launches Blackwell Ultra GPU: Doubles FP8 Performance, Achieves 16 TB/s HBM4 Bandwidth, Supports 100,000+ GPU Clusters for Trillion-Parameter AI Training
NVIDIA Corporate Newsroom USA Overview NVIDIA has unveiled its Blackwell Ultra GPU, engineered to meet the extreme demands of training trillion-parameter AI models. This new architecture delivers a 2x increase in FP8 performance and inte... -
Edge AI Shifts to Micro-Desktops & On-Device Personal Systems: 100 Billion Parameter Models Now Run Locally
Metaplugs News International Overview The AI industry is undergoing a fundamental hardware restructuring from cloud-only inference to compact on-device systems. A new generation of NPU-equipped mini-AI desktop PCs enables local execution... -
FriskAI Emerges from Stealth with $3.6 Million Pre-Seed Funding for AI Agent Runtime Monitoring Platform
Dealroom.co Unknown Overview FriskAI emerged from stealth with $3.6 million in pre-seed funding to develop a platform that monitors AI agents in production, specifically targeting regulated industries like healthcare, insurance, and fina... -
Open-Source LLMs Like Llama 3 Achieve Performance Parity with Proprietary Models in July 2026, Gaining Traction for Customizability and Flexibility
LLM Stats Unknown Overview In July 2026, open-source Large Language Models (LLMs) demonstrated remarkable progress, with models such as Llama 3, Mistral, Qwen, and DeepSeek matching or exceeding proprietary alternatives across numerous b... -
Moonshot AI Releases Kimi K3, a 2.8 Trillion Parameter Multimodal AI Model with a 1 Million Token Context Window, Slated for Open-Weight Release
Build Fast with AI China Overview China's Moonshot AI launched Kimi K3 on July 16, 2026, a 2.8 trillion parameter Mixture-of-Experts (MoE) multimodal AI model. This model features an unprecedented 1 million token context window and suppo... -
Former OpenAI CTO Mira Murati’s Thinking Machines Lab Unveils ‘Inkling,’ a 975-Billion Parameter Open-Weight Multimodal AI Model
BigGo Finance USA Overview Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, has introduced 'Inkling,' its first open-weight multimodal AI model. Inkling employs a 975-billion parameter Mixture-of-Experts (MoE) architectur... -
Moonshot AI Releases 2.8 Trillion Parameter Kimi K3 MoE Model, NVIDIA Introduces 4B-Parameter Cosmos 3 Edge for On-Device AI
AI Weekly USA Overview Moonshot AI officially released Kimi K3, a 2.8-trillion-parameter Mixture-of-Experts model with a 1-million-token context window and native vision capabilities. In parallel, NVIDIA introduced Cosmos 3 Edge, a 4-bil...
12