+- Inference (https://www.datasculpting.ai/inference)
+-- Forum: Harnesses, Agents & Orchestration (https://www.datasculpting.ai/inference/forumdisplay.php?fid=11)
+--- Forum: Workflows / Automation (https://www.datasculpting.ai/inference/forumdisplay.php?fid=47)
+--- Thread: DeepSeek V4.1 impressions (/showthread.php?tid=560)
Pages:
1
2
DeepSeek V4.1 impressions - rocmriley - 09-28-2026
Thought I'd post a small recipe because a lot of threads ask similar things.
This isn't revolutionary, but the combination of PyTorch and 4090 felt surprisingly usable once I tuned around MoE serving.
Feel free to tear it apart if you see something dumb.
This is mostly for local work rather than public serving.
Re: DeepSeek V4.1 impressions - fastquill74 - 09-29-2026
Tried to reproduce it just now. On 7900 XTX with Ollama, I saw basically the same behavior. I expected the model choice to matter more than it did.
Re: DeepSeek V4.1 impressions - routecraft - 09-30-2026
Thanks, that helps. That matches what I saw on mini PC. The biggest change for me was simplifying the stack instead of adding one more optimization layer. Turns out 'just one more test' was a lie.
Re: DeepSeek V4.1 impressions - quantghost - 10-03-2026
My first guess would be this: If you're on MLX, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.
Re: DeepSeek V4.1 impressions - routecraft - 10-03-2026
Quote:My first guess would be this: If you're on MLX, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of these
Tried to reproduce it just now. I couldn't reproduce the slowdown until I pushed context much higher. The surprise was that memory pressure showed up before throughput collapsed.
Re: DeepSeek V4.1 impressions - oxiderunner - 10-06-2026
Quote:[quote]My first guess would be this: If you're on MLX, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of
Maybe, but I had the opposite result. I wouldn't optimize around that first. On my side the bigger swing came from the runtime and not the model. Worth testing the simpler explanation before adopting the more interesting one.
I'm happy to test one more variant if you want a second data point.
Re: DeepSeek V4.1 impressions - routecraft - 10-06-2026
subscribed, curious where this lands.
Re: DeepSeek V4.1 impressions - promptpoint - 10-06-2026
Ran a quick check on my box. I couldn't reproduce the slowdown until I pushed context much higher. I expected the model choice to matter more than it did.
Re: DeepSeek V4.1 impressions - routecraft - 10-06-2026
Thanks, that helps. I ended up in about the same place using LM Studio. The biggest change for me was simplifying the stack instead of adding one more optimization layer.
Re: DeepSeek V4.1 impressions - routecraft - 10-09-2026
Appreciate the detail. That matches what I saw on mini PC. The biggest change for me was simplifying the stack instead of adding one more optimization layer.