Models & Labs
Google Releases Gemini Robotics 2 Software
Matt Wolfehigh confidence
Why it matters
- →Represents a major leap in humanoid robot capabilities.
- →Enables robots to perform complex tasks with precision.
- →Provides developers and integrators with advanced AI tools.
Google has unveiled Gemini Robotics 2, a significant advancement in AI software for humanoid robots. This release enhances coordination from feet to fingertips, allowing robots to balance, walk, and handle delicate tasks like sealing Ziploc bags and tying knots. The software includes 22 degrees of hand movement and can run locally without an internet connection. It adapts to new robot bodies within hours and powers hardware from partners such as Apptronik's Apollo 2. Developers can access the reasoning model 'Gemini Robotics ER 2' on Google AI Studio, while hardware integrators can access the full Vision-Language-Action and On-Device models through Google DeepMind's early access program.
Read originalThe story around this
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
More in Models & Labs
Models & Labsother
llama.cpp b12340 adds CUDA 13 and ROCm 10.0
This release quietly closes the hardware gap for local inference by adding default support for CUDA 13 and ROCm 10.0 across Linux and Windows. Users with newer NVIDIA GPUs or AMD accelerators no longer need to compile from source to get performance, while Apple Silicon builds now ship KleidiAI disabled by default, shifting the focus back to standard ARM optimizations. The inclusion of Snapdragon support for Linux and Android marks a significant step toward on-device AI beyond the x86/ARM desktop ecosystem. It is less about new features and more about ensuring llama.cpp remains the universal runtime for every available accelerator.
llama.cpp Releases
Models & Labscoding
llama.cpp b12342 adds few-row MMQ for speculative decoding
Speculative decoding just got a significant speed boost on NVIDIA GPUs. This release introduces a specialized kernel for 'few-row' matrix multiplications, targeting the small batch sizes typical of verification passes in speculative decoding workflows. By optimizing these specific operations for quantized formats like Q8_0 and IQ4 variants, llama.cpp reduces latency where it matters most during token generation. The change is CUDA-only for now but leverages hardware capabilities from Ampere onwards, making local inference faster without requiring new model architectures or complex configuration changes.
