The vLLM project has released version 0.22.0, featuring substantial improvements across its AI model infrastructure. This update includes 459 commits from 230 contributors, focusing on enhancing model performance and efficiency. Key advancements include the reorganization of the DeepSeek V4 model and the introduction of NVFP4 fused MoE support, which aim to improve accuracy and processing speed. The Model Runner V2 now defaults to Qwen3 dense models, enhancing performance with new features like sleep-mode weight reload. These updates position vLLM as a more robust framework for handling complex AI tasks.
Read originalThe latest release of llama.cpp, b10955, tackles a critical issue of heap corruption by disabling the ggml-cpu precompiled header and fixing CACHE_LINE_SIZE ambiguity. This update ensures consistent CACHE_LINE_SIZE values across C++ kernels and C work-buffer sizing code, preventing heap-buffer-overflow and subsequent crashes. By restoring the natural include order and removing the std::hardware_destructive_interference_size branch, the update makes the value deterministic and include-order independent. This release is a technical fix that stabilizes the runtime environment for developers using llama.cpp.
The latest llama.cpp release, b10956, introduces significant improvements to the SYCL backend, particularly for handling large k values in TOP_K operations. By implementing a radix select method, the update allows for efficient GPU-resident processing, avoiding previous limitations that forced operations to fall back to the CPU. This change enhances performance, especially in scenarios requiring large k values, such as qwen4exp's sparse-attention indexer. The update ensures that operations are more efficient and scalable, providing a notable boost in processing speed without regressing any measured shapes.