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.
RTPurbo, a technique that converts a Full Attention LLM into a highly efficient Sparse model with only a few hundred short training steps, has been unveiled.
Elon Musk announced that a Grok model with 0.5T parameters will be released next year.
Google has shifted its scientific AI strategy by announcing 'Gemini for Science,' centered on LLM-based agents instead of specialized models.
Uber's COO pointed out the cost-efficiency problem arising from maximizing LLM token usage.
An interview with Greg Brockman covering OpenAI's technical turning points and the structural changes needed to achieve AGI.
Antigravity introduces Gemini 3.5 Flash (Low) for token optimization and resets quotas for paid plans.
This clearly defines easily confused concepts like Harness, Scaffold, and Model when building AI Agents.
The CC 2.1.147 update adds code review agent features supporting various difficulty and verification levels.
Introducing DeepSeek reasonix, a coding agent that leverages DeepSeek models to boost caching efficiency and cut costs.
Anthropic is preparing a Claude memory update based on 'Memory Files' that stores information structured by topic.
Huawei unveiled the BitCPM-CANN model series, which supports 1.58-bit quantization-aware training on Ascend NPUs.
DeepSeek is permanently cutting the price of its flagship AI model by 75%.
Anthropic's new model 'Mythos' appears set to be integrated into Claude Code and Claude Security.
This analyzes various AI hardware technologies and corporate strategies aimed at solving the memory bandwidth problem, the core bottleneck in LLM inference.
An MLX port has been implemented that allows Cohere's 218B MoE model, Command A+, to run on Apple Silicon.
DeepSeek has confirmed the 75% discounted price for the V4-Pro API as the permanent official price.
DeepSeek has permanently cut DeepSeek-V4-Pro API prices to 1/4 of the previous level and significantly lowered cache hit prices as well.
Google Antigravity 2.0 and Gemini 3.5 Flash High achieved the highest quality scores in an OpenSCAD-based architectural 3D modeling benchmark.
DeepSeek announced that it is permanently applying a 75% discount to v4 Pro model pricing.
Google unveiled Gemini 3.5 Flash, a high-speed model optimized for agentic tasks.
Norway has deployed 2PB of Huawei flash storage for LLM training.
DeepSeek plans to secure investment of approximately $10.29 billion and focus on open-source AI development and achieving AGI.
Cohere has released Command A+, an Apache 2.0-licensed model featuring lossless quantization and native citation capabilities.
AI scrapers are heavily shaking up the costs and stability of public wikis.