Ollama has released version 0.18, bringing significant improvements for developers building agentic AI applications. The release includes official OpenClaw provider integration, performance improvements for cloud models, and the introduction of NVIDIA Nemotron-3-Super.
OpenClaw Integration
Ollama can now be selected as an authentication and model provider during OpenClaw onboarding. This integration bridges the gap between local LLM inference and agentic AI workflows.
openclaw onboard --auth-choice ollamaPerformance Improvements
Ollama 0.18 delivers up to 2x faster speeds with Kimi-K2.5, along with improved tool calling accuracy. For North American users, MiniMax-M2.5 and Qwen3.5 on Ollama cloud now respond up to 10x and 2x faster respectively.
Nemotron-3-Super
The highlight is Nemotron-3-Super, a 122B parameter model with strong reasoning and tool calling. It scores highest of any open model on PinchBench, measuring task completion success with OpenClaw.
- Cloud:
ollama run nemotron-3-super:cloud - Local:
ollama run nemotron-3-super(requires 96GB+ VRAM)
Non-Interactive Task Support
The ollama launch command now supports --yes for non-interactive use in scripts and CI/CD:
ollama launch claude --model glm-5:cloud --yes -- "Review this PR"