Coding Tools
llama.cpp v0.5.0: Backend performance and multi-address server
llama.cpp Releaseshigh confidence
Why it matters
- →CUDA conv2d acceleration via implicit GEMM improves throughput for vision-language models.
- →Multi-address server binding enables more flexible routing in distributed inference setups.
- →Metal MoE and SSM fusion optimizations boost performance on Apple Silicon hardware.
llama.cpp has released version 0.5.0, focusing on backend performance improvements and server robustness. Key technical updates include CUDA conv2d acceleration using implicit GEMM and Metal MoE/SSM fusion optimizations to enhance throughput on Apple Silicon. The release also adds support for binding the server to multiple addresses, improving deployment flexibility for distributed inference. New model coverage includes HRM-Text (DFM Mimir 1B), MiMo-V2.6, and Qwen4Exp, alongside updates to the ggml backend v0.25.0.
Read originalMore from llama.cpp Releases
Coding Toolscoding
llama.cpp b11140 optimizes sparse attention on CUDA
This release targets a specific bottleneck in long-context inference by optimizing the sparse flash attention prefill step for NVIDIA GPUs. By templating kernels to unroll loops at compile time, batched sparse operations drop from 586 microseconds to 244 microseconds on 49k context windows. This isn't just a generic speed bump; it makes handling very long documents significantly more efficient for users relying on sparse attention mechanisms. The change is already baked into the standard CUDA builds, requiring no special flags.
llama.cpp Releases
Coding Toolscoding
llama.cpp b11146 adds CUDA 13 and Snapdragon support
The latest llama.cpp build brings immediate relevance to users on bleeding-edge NVIDIA hardware with native CUDA 13.4 support across Linux and Windows, closing the gap for those testing next-gen GPU architectures. More notably, it finally addresses the mobile inference landscape by including a dedicated build for Linux arm64 Snapdragon devices, covering CPU, Adreno GPU, and Hexagon NPU paths. This moves local AI beyond just desktop GPUs into the realm of high-performance edge computing on Qualcomm silicon. While Apple Silicon builds have KleidiAI disabled in this specific release, the expansion to ARM-based mobile NPUs marks a significant shift in where llama.cpp can run efficiently.
More in Coding Tools
Coding Toolscoding
Claude Code v2.1.281 fixes session stability and adds Bedrock guardrails
This release stabilizes Claude Code's core reliability by fixing persistent bugs in session resumption and prompt caching that previously caused data loss or infinite loops. It also tightens enterprise security with new Bedrock upstream support for IAM role assumption and mandatory guardrail application. The update addresses critical edge cases like proxy stream drops and oversized tool calls, ensuring smoother operation in complex development environments.
Claude Code Releases
© Duncan RogoffCoding Toolscoding
Anthropic releases official Claude Code playbook for Opus 5.5
Anthropic has published an official guide on optimizing Claude Code with the new Opus 5.5 model, shifting focus from raw capability to engineered workflow. The playbook details specific prompting habits that reduce latency and cost while increasing autonomy during long coding sessions. This moves beyond generic advice, offering concrete strategies for developers who rely on agentic coding tools for complex tasks. It signals a maturation in how enterprise-grade AI assistants are integrated into daily engineering practices.
