Local Ai
v0.20.4-rc2: gemma4: Disable FA on older GPUs where it doesn't work (#15403)
Ollama v0.20.4-rc2 is a release candidate that addresses a compatibility issue with Flash Attention (FA) for the Gemma 4 model on older GPUs. CUDA versions older than 7.5 lack the support needed t...
Ollama v0.20.4-rc2 is a release candidate that addresses a compatibility issue with Flash Attention (FA) for the Gemma 4 model on older GPUs. CUDA versions older than 7.5 lack the support needed to enable Flash Attention for the model , so this patch disables FA for those affected hardware configurations. The fix was tagged by dhiltgen on April 7, 2026, and preceded the full v0.20.4 stable release, which formally enabled Flash Attention for Gemma 4 on supported hardware.
Related
- v0.20.4
- v0.20.5-rc2
- v0.20.5-rc0
- v0.20.5
- Using Ollama Gemma4 models via OpenWebUI on my phone and it’s been a good experience
- OpenClaw + Ollama + gemma4:26b is fast in raw Ollama, but first heavy OpenClaw turns are extremely slow or hit idle timeout
Source: local-ai