Ollama v0.34.3 Adds Model Thinking Controls and Nemotron Vision Support

Ollama's latest release introduces fine-grained control over model thinking/reasoning behavior through the API, allowing developers to tune inference characteristics at deployment time. The addition of thinking level controls (low, high, max) with advertised defaults enables more sophisticated local inference patterns where reasoning depth can be adjusted based on task requirements and hardware constraints.

The expanded Nemotron vision model support on Apple Silicon represents progress in the multimodal inference space for local deployments. These framework improvements demonstrate the ongoing maturation of local LLM infrastructure, moving from basic text generation toward more nuanced inference control and broader model family support across consumer hardware platforms.

Read the full article on Ollama release.

Source: Ollama release · Relevance: 8/10