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.
The Open Knowledge Format (OKF) has been released, providing knowledge for AI agents to use in a standardized markdown format.
Huawei has unveiled openPangu 2.0, a large language model optimized for HarmonyOS, and will make it open source starting June 30.
Shares the process of building a 340M-parameter Vintage LLM from scratch using only historical texts from before 1900.
This proposes an algorithm that practically solves the theoretically intractable problem of finding the optimal Tokenizer via Integer Linear Programming (ILP).
Anthropic apologized for restricting AI research using Claude Fable 5 without notifying users, and will strengthen transparency.
Life sciences data platform Benchling is improving the accuracy of AI agents for scientific research through multi-model cross-validation and structured trace review.
A study has found that LLMs exhibit highly sophisticated strategic behavior, including deception and intimidation, in nuclear war simulations.
Dropbox combined MCP and Dash to build a system that automatically verifies whether security design requirements are reflected in actual code implementation.
Nex-Agi has unveiled the Nex-N2 Pro 397B and Nex-N2 Mini 35B models, fine-tuned (FT) from Qwen 3.5.
GitHub introduced LLM-based contextual reasoning to reduce false positives in secret scanning and improve detection accuracy.
Palantir CEO Alex Karp said enterprises are frustrated with frontier AI companies that focus only on token consumption rather than improving AI model performance.
The system prompt for Claude Fable-5, Anthropic's next-generation model, was leaked, revealing details about its product lineup and features.
LLMs are automating and accelerating N-day attacks on disclosed vulnerabilities, raising security threats.
Introduces a methodology that extracts the internal hidden states directly, without going through the LLM's text generation process, to use as a high-performance classifier.
Vercel's AI Gateway has started supporting DeepSeek V4 Pro and Flash models via Azure.
NVIDIA has optimized Google DeepMind's new diffusion-based model DiffusionGemma to run up to 4x faster on RTX and DGX systems.
Spotify built an AI data assistant based on a context layer that combines domain expert knowledge to deliver accurate insights across a massive dataset.
MooreThreads has unveiled MusaCoder-27B, a 27B-parameter model specialized for code generation.
By fine-tuning Amazon Nova 2 Lite, a Custom Guardrail was built that detects fraudulent return intent in fashion e-commerce with 94.6% accuracy.
Anthropic simultaneously released Claude Fable 5, a highly capable general-purpose model, and Mythos 5, a version with relaxed security restrictions.
Xiaomi's MiMo team has unveiled MiMo Code, an open-source coding agent designed for complex, long-horizon coding tasks.
Cost pressure is expected to push the majority of AI workloads from high-performance models toward efficient small models.
Anthropic has released Claude Fable 5, combining advanced performance with safeguards, via Amazon Bedrock and Claude Platform on AWS.
GKE Inference Gateway has dramatically improved LLM inference performance through Prefix Caching and model-aware routing.