LLM
Tencent's 440 MB AI model translates 33 languages offline on your phone
Tencent released a compact 440 MB AI translation model as an open-weight model that supports 33 languages and runs...
The Decoder
LLM · Other Companies
LLM
IBM Releases Two Granite Speech 4.1 2B Models: Autoregressive ASR with Translation and Non-Autoregressive Editing for Fast Inference
IBM has released two versions of the Granite Speech 4.1 2B model for automatic speech recognition (ASR), featuring both...
MarkTechPost
LLM
Qwen Team Releases FlashQLA: a High-Performance Linear Attention Kernel Library That Achieves Up to 3× Speedup on NVIDIA Hopper GPUs
The QwenLM team released FlashQLA, a high-performance kernel library that accelerates linear attention mechanisms,...
MarkTechPost
LLM
Sanctioned Chinese AI Firm SenseTime Releases Image Model Built for Speed
SenseTime, a sanctioned Chinese AI firm, has released a new image generation model optimized to run on Chinese-made...
Wired AI
LLM
With Nemotron 3 Nano Omni, Nvidia reveals what really goes into a modern multimodal model
Nvidia releases Nemotron 3 Nano Omni, an open multimodal model capable of processing text, image, video, and audio. The...
The Decoder
LLM
Here is what an LLM that knows nothing after 1930 thinks our world looks like in 2026
A 13-parameter language model called 'Talkie' trained exclusively on texts from before 1931 generates predictions for...
The Decoder
LLM
OpenMOSS Releases MOSS-Audio: An Open-Source Foundation Model for Speech, Sound, Music, and Time-Aware Audio Reasoning
OpenMOSS released MOSS-Audio, an open-source foundation model that unifies speech, sound, music, and temporal reasoning...
MarkTechPost
LLM
The company with a monopoly on AI's most critical machine is racing to build more
ASML, which holds a monopoly on EUV lithography machines essential for AI chip production, is significantly increasing...
The Decoder
LLM
The LoRA Assumption That Breaks in Production
LoRA, a popular efficient fine-tuning method for large models, relies on the assumption that all model updates are...
MarkTechPost
LLM
How to Build a Fully Searchable AI Knowledge Base with OpenKB, OpenRouter, and Llama
This tutorial demonstrates how to build a searchable AI knowledge base using OpenKB with Llama models accessed through...
MarkTechPost
LLM
500 investment bankers review AI outputs and find none ready for client delivery
A benchmark study where 500 investment bankers evaluated outputs from leading AI models like GPT-5.4 and Claude Opus...
The Decoder
LLM
A Coding Implementation on kvcached for Elastic KV Cache Memory, Bursty LLM Serving, and Multi-Model GPU Sharing
This tutorial explores kvcached, a dynamic KV-cache implementation for vLLM that optimizes GPU memory usage in large...
MarkTechPost
LLM
Qwen3.6-27B beats much larger predecessor on most coding benchmarks
Alibaba's new open-source model Qwen3.6-27B with 27 billion parameters outperforms its significantly larger 15x...
The Decoder
LLM
Xiaomi Releases MiMo-V2.5-Pro and MiMo-V2.5: Matching Frontier Model Benchmarks at Significantly Lower Token Cost
Xiaomi's MiMo team released two new open-source models, MiMo-V2.5-Pro and MiMo-V2.5, that achieve frontier model...
MarkTechPost
LLM
Alibaba Qwen Team Releases Qwen3.6-27B: A Dense Open-Weight Model Outperforming 397B MoE on Agentic Coding Benchmarks
Alibaba's Qwen Team released Qwen3.6-27B, a 27-billion-parameter dense open-weight model that outperforms a 397B MoE...
MarkTechPost
LLM
Teaching AI models to say “I’m not sure”
A new training method enables AI models to better estimate their own confidence levels and acknowledge uncertainty,...
MIT News AI
LLM
AI in law firms entering its closing summaries
Paris-based AI consultant Olivier Chaduteau describes three phases of AI adoption in law firms: initial dismissal,...
AI News
LLM
The flood of AI music is reshaping how streaming platforms handle new uploads
Deezer reports that 44 percent of daily song uploads to its platform are now fully AI-generated, prompting the...
The Decoder
LLM
A Coding Implementation on Qwen 3.6-35B-A3B Covering Multimodal Inference, Thinking Control, Tool Calling, MoE Routing, RAG, and Session Persistence
This tutorial demonstrates a practical implementation of Qwen 3.6-35B-A3B, a multimodal MoE model, covering key...
MarkTechPost