Ollama v0.31.2-rc2 enables iGPU multimodal projector offload with fit padding

Core summary

One sentence to understand this update

Ollama's latest release candidate allows multimodal projector offloading to integrated GPUs using "fit padding" for improved performance and compatibility.

Impact & opportunity

What this could mean

Builders can now leverage iGPUs for multimodal models in Ollama, potentially expanding local deployment options and performance for users without dedicated GPUs.