Ollama v0.40.7 Release: OpenAI API Improvements and MLX Enhancements

Ollama's v0.40.7 release brings important enhancements to OpenAI API compatibility and MLX support, addressing real-world deployment needs. The addition of Responses effort support on non-thinking models expands flexibility for teams using the OpenAI-compatible endpoint, while improvements to draft speculative decoding in llama-server enable faster inference through more effective model speculation strategies.

The MLX enhancements, including support for PPLX Decider v1.2, demonstrate Ollama's commitment to optimizing Apple Silicon performance. As MLX continues evolving and new model variants emerge, keeping Ollama aligned with these developments ensures Mac users get optimal performance from their hardware investments. These incremental improvements represent the ongoing maturation of Ollama as a production inference platform.

For teams already using Ollama in production or evaluating it for deployment, this release signals continued investment in API compatibility, performance optimization, and model support. The focus on both OpenAI-compatible endpoints and platform-specific optimizations (like MLX) suggests Ollama is becoming the unified inference platform for diverse local deployment scenarios, from cloud instances to edge devices.

Read the full article on Ollama release.

Source: Ollama release · Relevance: 8/10