b8112
The latest release patches a bug where assistant messages with both content and thinking caused failures.
Deep Dive
The ggml-org team behind the popular open-source project llama.cpp has released version b8112. This update fixes a critical Jinja template error in the 'gpt-oss' component that occurred when an AI assistant's message contained both standard content and internal 'thinking' alongside tool/function calls. The fix ensures greater stability for developers building complex AI agents that utilize reasoning and external tool integration, preventing crashes in multi-step workflows.
Why It Matters
This patch is essential for developers creating advanced AI agents that combine reasoning, tool use, and natural language output without errors.