DeepSeek4 - blog.raem.io

Llama.cpp Build b10217 Introduces Tool Calling in Thinking Mode for DeepSeek (DS4)

Leave a Comment / AI & Machine Learning / AI-Curated Insights

The latest llama.cpp release, build b10217, enables tool calling during the thinking phase for DeepSeek (DS4) models in chat mode. This enhancement allows the model to seamlessly invoke external tools while it is still reasoning, producing more coherent and effective multi-step interactions. For self-hosted users running DeepSeek, this means smarter assistant behavior and improved reliability […]

Llama.cpp Build b10217 Introduces Tool Calling in Thinking Mode for DeepSeek (DS4) Read More »

llama-arch Fixes DeepSeek4 APE Tensor Operation for Improved Model Compatibility

A recent update to llama-arch, a core component of the llamacpp inference engine, addresses a critical bug affecting the APE (Adaptive Positional Encoding) tensor operation used in DeepSeek4 models. This fix ensures accurate computation during model inference, which is essential for developers running DeepSeek4 on self-hosted environments. By resolving this issue, llama-arch improves compatibility and