v0.20.5-rc0

Ollama v0.20.5-rc0 is a release candidate for the v0.20.5 version of the Ollama open-source project, which enables users to run large language models locally. The stable v0.20.5 release that followed introduced features including OpenClaw channel setup (connecting WhatsApp, Telegram, and Discord via ollama launch openclaw), flash attention support for Gemma 4 on compatible GPUs, and a fix for the /save command for models with safetensors-based architectures. As an rc0 tag, this release served as a pre-release build for testing prior to the final v0.20.5 publication on April 9, 2026.

Related

• v0.20.5-rc2
• b8753
• v0.20.5
• b8750
• v0.20.4-rc2: gemma4: Disable FA on older GPUs where it doesn't work (#15403)
• v0.20.6

Source: local-ai