Models & Labs
Google DeepMind Launches Gemini 3.6 Flash Models
Google DeepMindhigh confidence
Why it matters
- →The Gemini 3.6 Flash series enhances token efficiency, reducing costs for AI agents.
- →3.5 Flash-Lite offers high throughput, crucial for scaling agentic workflows.
- →3.5 Flash Cyber addresses cybersecurity needs, improving vulnerability detection and response.
Google DeepMind has introduced the Gemini 3.6 Flash series, which includes the 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber models. The 3.6 Flash model improves token efficiency by 17% and reduces costs, enhancing coding and knowledge work. The 3.5 Flash-Lite model is optimized for high throughput tasks, offering 350 output tokens per second. Meanwhile, the 3.5 Flash Cyber model is tailored for cybersecurity, detecting and fixing vulnerabilities efficiently. These models are now available for developers and enterprises, promising advancements in AI agent performance and cybersecurity.
Read originalMore in Models & Labs
Models & Labsmodels
llama.cpp b10156 Release Expands Platform Support
The latest b10156 release of llama.cpp continues its trend of broadening platform compatibility, notably adding support for ROCm 7.2 on Ubuntu x64. This update ensures that AMD GPU users can leverage llama.cpp more effectively, narrowing the gap with NVIDIA's CUDA. The release also includes Vulkan support for both Ubuntu and Windows, enhancing the versatility of the software for developers. While no new models or quantization methods are introduced, this update solidifies llama.cpp's position as a versatile inference runtime across diverse hardware configurations.
llama.cpp Releases
Models & Labsmodels
Llama.cpp b10164 Release Enhances CUDA Performance
The latest b10164 release of llama.cpp focuses on improving CUDA performance, particularly for Mamba-2 prefill acceleration. By introducing chunked SSD matrix multiplication, the update aims to enhance efficiency and memory coalescing. This release also addresses several technical fixes, including resolving a read-write race condition in CUDA operations. While there are no groundbreaking new features, these optimizations make llama.cpp a more robust choice for developers working with CUDA and related technologies.