AI Briefing#llm
The latest AI and developer news about #llm, with the original source and a short summary.
The latest AI and developer news about #llm, with the original source and a short summary.
In an M3 Pro benchmark, Nemotron 3 Nano took first place among 4B-class models.
DeepSeek has lowered its API prices by up to 90% and also cut V4 costs by 75%.
Xiaomi has unveiled MiMo-V2.5-Pro, a 1.02T-parameter MoE model.
The HTX301-based PCIe card supports 700B LLM inference at about 240W.
TuneForge bundles data generation, LoRA fine-tuning, and evaluation together inside chat.
OpenAI announced it is stopping SWE-bench Verified reporting and shifting to new evaluation axes.
It explains the principles of vector quantization in TurboQuant and its random rotation-based compression method.
Explains why the MoE architecture lowers inference costs for DeepSeek's large-scale model.
The US State Department has instructed diplomats worldwide to publicize allegations of IP theft by Chinese AI companies including DeepSeek.
A VRAM estimation tool that runs llama.cpp's fit algorithm in the browser has been released.
A 23-year-old amateur solved a 60-year-old Erdős problem using ChatGPT Pro's GPT-5.4 Pro.
Stash has unveiled a persistent memory layer and MCP integration for AI agents.
Released an AI memory that achieved Recall@5 59% on the LoCoMo-10 benchmark.
Google will invest up to $40 billion in Anthropic through a combination of cash and compute.
A wiki system maintained directly by agents, based on Markdown and Git, has been released.
New pricing policies from OpenAI and DeepSeek have polarized the AI cost structure.
A technical analysis alleges that HauhauCS's abliteration package plagiarized the Heretic project and violated its license.
ElevenLabs shared a deployment and security guide for AI Agents in financial services.
Allegations have emerged of an OpenAI connection to a news site run by AI bot journalists.
A decensored version of Qwen3.6 35B A3B using Heretic has been released.
Trained a Reddit summarization model with GRPO on a 3x Mac Mini cluster, and quality+length reward achieved the best performance.
Pocket LLM v1.5.0 has been released, adding voice and image input, OCR, and camera capture to Android offline chat.