Muse Spark on M3U 256 GB
Instead of counting 'hmm' on DeepSeek v4.1 Flash 1017, I today clamped Muse Spark Open 816B-A42B to ds4.
Proof1 community thread
Key points
- This was the first successful run, FTW, that made it to decode...
- I'm hoping it will 4-5x with any luck.
- I'll obvs be running larger prefill chunks too (pre-metal leftovers), but I'm not expecting such leaps.
Sources (1)
- [1]Muse Spark on M3U 256 GBr/LocalLLaMA (top, daily) · Oct 18, 01:35 AM
Instead of counting 'hmm' on DeepSeek v4.1 Flash 1017, I today clamped Muse Spark Open 816B-A42B to ds4.
This was the first successful run, FTW, that made it to decode...
Extractive summary: sentences quoted from the sources.
Before this
- Oct 18, 2026DeepSeek V4.1 Flash 1017 on 2x DGX Spark: 57 tok/s text, 92 tok/s code with TensorFold (2.1-2.3x the vLLM kit)
- Oct 17, 2026DeepSeek V4.1 Flash 1017 vs V4.1 Flash, Pi vs DeepSeek Harness: results on my 10 coding tasks
- Oct 17, 2026DeepSeek V4.1 Flash 1017 token efficiency
- Oct 17, 2026deepseek-ai/DeepSeek-V4.1-Flash-1017
- Oct 17, 2026Nemotron 3.5 Bolt, Muse Spark Open, and a Fast LFM2.5 24B
- Oct 15, 2026model: add Muse Spark Open (text, vision and audio) by ServeurpersoCom · Pull Request #30616 · ggml-org/llama.cpp
