AI Intel

Expand Board for macOS, a new tool launched on Product Hunt, allows users to visualize and organize ideas, concepts, and plans across infinite digita…

Nvidia has announced CUDA Rust, enabling native GPU programming directly within the Rust language, offering two tracks for writing GPU kernels.

New research explores methods to break the 1.58-bit barrier for Ternary LLMs, pushing the boundaries of efficient model quantization.

An experiment ran Qwen 3.8 27B 4-bit quantized with a 100K context window for 63 hours on an RTX 3090 in an attempt to solve the Riemann Hypothesis,…

New research proposes optimal model activation policies for inference networks of large language models, aiming to address the high inference costs o…

ZGCM-1 is a new 7B dense foundation model, fully open and highly efficient, specifically trained for math and agentic search capabilities from scratc…

Weave Router 2.0 在 Product Hunt 发布，作为一款支持订阅感知的编码智能体路由器。

Ollama 更新至 v0.34.1，显著增强了 Apple Silicon 上的 MLX 内存处理能力，`ollama create` 功能不再处于实验阶段，并改进了 GGUF 模型创建流程。

Anthropic 的 Claude Code API 发布 v2.1.273，新增多项 `x-claude-code` 请求头，支持 LLM 网关进行更精细的 Agent 类型、工具时长、压缩及上下文管理。

llama.cpp 发布 b10999 版本，主要改进了统一 KV 缓存的最大上下文长度自动适配功能。