DEV Community

Discussion on: Running Gemma 4 Locally with Ollama and OpenCode

Collapse
 
grovertek profile image
GroverTek

I'm curious which provider you are using to access Gemma 4 for free in the cloud. My own searching hasn't yet found a cloud source.

My experience with Ollama is a little spotty for more complex tasks. Gemma, Qwen 3.5, etc. have a tendency to work great for a bit and then they just stop mid task. Something about the tool hand off situation. The fix is to refine the prompt to get more standard responses, but that is not always feasible when you are in the middle of a loop over thousands of records and it breaks in the middle of the night while unattended. This means I can't trust the Ollama approach for long running tasks, at least not yet. I'm assuming this is mostly a skills or configuration issue on my end though. I've found though with Big Pickle and other models I can run long tasks, as long as I'm not hammering the cloud server too hard and get rate limited. If I can use Gemma 4 in the cloud for free, that may change my outlook. (I don't like paying multiple subscriptions for LLMs, and it adds up too quickly for me)

Collapse
 
shemshallah profile image
Justin Howard-Stanley
Thread Thread
 
grovertek profile image
GroverTek

Thanks. I understood Ollama's cloud system was a subscription, but never really looked into too deeply. I'll check that out in a bit. In the mean time I did find a free option on OpenRouter: openrouter.ai/google/gemma-4-31b-i...