Ollama 发布新模型调度系统，改为精确测量显存
New model scheduling · Ollama Blog
Ollama 上线新的模型调度系统，运行前会精确测量所需显存，取代此前版本的估算方式。官方称这能显著减少显存不足导致的崩溃、把更多显存分配给 GPU 以提升 token 生成与处理速度，并改善多卡及不同型号 GPU 混用时的调度表现，nvidia-smi 与 ollama ps 的显存读数也将一致。新引擎中的模型默认启用该功能，已支持 gpt-oss、llama4、gemma3、qwen3、mistral-small3.2 等，llama3.2、llama3.1、qwen3-coder 等将陆续迁移。
正文
当前显示原文，中文全文译文尚未就绪。
September 23, 2025
Ollama now includes a significantly improved model scheduling system. Ahead of running a model, Ollama’s new engine will now measure the exact amount of memory required compared to an estimation in previous versions of Ollama. This has several benefits:
Significantly reduced crashes due to out of memory issues: Because memory management is exact, over-allocations no longer occur meaning fewer out of memory issues; Maximizing GPU utilization: Ollama’s new memory management allocates more memory to the GPU, increasing token generation and processing speeds; Multi-GPU performance: Ollama will now schedule models more efficiently over multiple GPUs, significantly improving multi-GPU and mismatched GPU performance; Accurate reporting: Measurements in tools like nvidia-smi will now match ollama ps making it easy to track memory utilization on your system
All models implemented in Ollama’s new engine now have this new feature enabled by default, with more models coming soon as they transition to Ollama’s new engine.
Examples
Long context
GPU: 1x NVIDIA GeForce RTX 4090; Model: gemma3:12b; Context length: 128k
Image input
GPU: 2x NVIDIA GeForce RTX 4090; Model: mistral-small3.2; Context length: 32k
Supported models
All models implemented in Ollama’s new engine use the new memory management features:
gpt-oss; llama4, llama3.2-vision (soon: llama3.2, llama3.1, llama3); gemma3, embeddinggemma, gemma3n; qwen3, qwen2.5vl (soon: qwen3-coder); mistral-small3.2; all-minilm and other embedding models