Hugging Face and Pollen Robotics demonstrated a fully local conversational AI pipeline on the Reachy Mini desktop robot, using Silero VAD v5, Parakeet-TDT 0.6B v3, Gemma 4 or Qwen3-4B LLM, and Qwen3-T…
A new benchmark test of Ornith 1.0, a model that builds its own task scaffolds, found that providing a full shell and Python environment doubled its bug-finding performance without increasing false po…
Developer Rand01ph released GemmaTrans, an on-device translation app for macOS built with MLX-Swift, supporting Google Gemma 4 and Tencent Hy-MT2 models fully offline. The app provides a local HTTP AP…
A DigitalOcean benchmark found that serverless inference providers produce different LLM behavior across deployments, with provider rankings flipping by model and diverging on speed, output fidelity, …
DeepReinforce released Ornith-1.0, an open-source family of coding models that learn their own reinforcement learning scaffolds, achieving state-of-the-art results among open models. The lineup includ…
Ornith-1.0, a family of open-source LLMs specialized for agentic coding, achieves state-of-the-art performance on multiple coding benchmarks including SWE-Bench and Terminal-Bench. The models range fr…
DeepReinforce open-sourced Ornith-1.0, a family of self-improving coding models ranging from 9B to 397B parameters, which learn to generate their own task-specific scaffolds during reinforcement learn…
Flama 2.0 introduces a CLI tool that allows users to download, package, and serve large language models from HuggingFace with a single command, including a built-in chat interface and production-ready…
A developer participating in the Gemma 4 challenge built Dresscode, an AI stylist that uses computer vision to digitize wardrobes and LLM function calling to recommend weather-appropriate outfits. The…
Google's Gemma 4 31B outperforms Alibaba's Qwen 3.6 27B on agentic code review tasks, finishing faster due to superior Multi-Token Prediction (MTP) design, according to benchmarks and field reports. W…
A critical bug in Google's Gemma 4 causes it to malform tool calls under real load, affecting vLLM, llama.cpp, Ollama, and oobabooga. A developer open-sourced a diagnosis, repair, and experimental LoR…
Qwen's coder outperforms Gemma 4 by 21 points on a benchmark, but the lead narrows significantly when both models run on a 16GB budget against a real repository, suggesting the bottleneck lies elsewhe…
Google Cloud and Anyscale announced optimizations for Ray Serve LLM on Google Kubernetes Engine (GKE) that deliver up to 5x higher throughput and 8x lower latency for large language model inference. T…
Fable 5, an AI agent, achieved 255 tokens per second on Gemma 4 inference using WebGPU before its access was suspended. The developer released the demo and kernels, claiming agentic kernel optimizatio…
Google engineer Rody Davis demonstrated how AI agents can achieve 100X engineering productivity using Antigravity 2.0, an agent-first platform that unbundles development tools into four pillars: Agent…
A new Hugging Face Space demonstrates Gemma 4 E2B running in-browser via WebGPU at 255 tokens per second, showcasing efficient on-device AI inference.…
Researchers at Google DeepMind instrumented DiffusionGemma 26B, a masked discrete-diffusion mixture-of-experts model, and found its decoding is neither parallel nor sequential but follows a partial le…
ML engineer Vicki Boykis's blog post 'Running Local Models is Good Now' hit the top of Hacker News, signaling that local AI models have crossed from hobbyist experiment to genuine developer workflow i…
Google Cloud Platform (GCP) provides a multi-layered approach to DNS abuse mitigation using Cloud DNS, Cloud Armor, and the Gemma 4 AI model. Cloud DNS offers managed DNSSEC to prevent cache poisoning…
A developer built Solstice Eternal, an AI-powered adventure game using Gemma 4. The game features interactive storytelling, dynamic decision-making, and a responsive web interface, with Gemma 4 assist…