RSS Reader

OpenAI 發布全新推理模型 o3 及 o4-mini，首次實現圖像推理及多工具整合，數學及編程表現顯著提升，引發全球科技界關注。此舉標誌 AI 技術向「代理型 AI」邁進，具備自主決策及跨模態問題解決能力。OpenAI 將以o3 取代o1模型，

Open Source LLM

微軟發表首個超過20億參數的1-bit模型同樣效能但更省電、不佔記憶體

ITHome | April 17, 2025, 12:04 a.m. | Original

微軟本週發表20億參數的1-bit模型BitNet b1.58 LLM家族，稱此新型模型比主流Transformer LLM更不佔記憶體且更少耗能，適合在CPU或較小型硬體平台上執行。

微軟研究院與中國科學院研究人員2023年發表名為《BitNet: Scaling 1-bit Transformers for Large

Open Source LLM

TurnkeyML 6.2 Released With AMD Ryzen AI NPU Improvements

Phoronix | April 16, 2025, 8:22 p.m. | Original

Released at the end of 2023 was TurnkeyML as an open-source collaboration between ONNX and AMD developers . TurnkeyML has evolved into focusing on making it easy to use the most important tools within the ONNX ecosystem and their Lemonade SDK to deploy large language models on various devices/accelerators Out today is TurnkeyML 6.2 with a focus on delivering better AMD Ryzen AI NPU support.

With today's TurnkeyML 6.2 release the Ryzen AI NPU support is now available as a selectable device backend within the lemonade

Open Source LLM

New approach from DeepMind partitions LLMs to mitigate prompt injection

TechSpot News | April 16, 2025, 4:43 p.m. | Original

Serving tech enthusiasts for over 25 years.
TechSpot means tech analysis and advice you can trust .

In context: Prompt injection is an inherent flaw in large language models, allowing attackers to hijack AI behavior by embedding malicious commands in the input text. Most defenses rely on internal guardrails, but attackers regularly find ways around them – making existing solutions temporary at best. Now, Google thinks it may have found a permanent fix.

Since chatbots went mainstream in 2022, a security flaw known as prompt injection has

Open Source LLM

Qt Ramps Up Its AI Coding Assistant Capabilities

Phoronix | April 16, 2025, 6:20 a.m. | Original

Over the past two years Qt has been working to embrace generative AI for enhanced coding capabilities within Qt Creator for Qt/QML/C++ applications. It started with GitHub Copilot integration and has been continuing well beyond that and today marks the debut of the Qt AI Assistant v0.9.

The Qt AI Assistant v0.9 release is designed to provide the latest AI-driven coding help for constructing Qt/C++ software. There is now local LLM support with this Qt AI Assistant via the

Open Source LLM

OpenAI Is Reportedly Developing An X-Like Social Media Platform Built Around ChatGPT’s AI Image Capabilities

wccftech | April 15, 2025, 9:39 p.m. | Original

While OpenAI has been completely focusing on expanding and extending its technology to wider markets by constantly introducing new models and updates, the competition in the industry is getting fierce, especially with Chinese players grabbing quite the attention. After DeepSeek's growing popularity, the ChatGPT maker has become more aggressive with its approach. It recently introduced GPT-4.1 and is working on advanced reasoning models. Now, OpenAI is reportedly working on a new social media platform that is said to be similar to X and mirror the same format

Open Source LLM

OpenAI Releases GPT-4.1 Model Family with Massive Context and Improved Performance

TechPowerUp News | April 15, 2025, 12:48 p.m. | Original

Yesterday, OpenAI unveiled its new GPT-4.1 model family, which includes GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano. These models promise significantly improved coding, instruction following, and long-context understanding compared to earlier versions like GPT-4o. All three support up to one million tokens of context, so they can handle much larger documents, codebases, or even hour-long video transcripts. In coding tests, GPT-4.1