⚡ Update: v2 (post #71) achieves 51 tok/s. v2.1 (post #104) adds a quick-start script. See those posts for the latest setup. Been chasing every last token/second out of Qwen3.5-122B-A10B on a single DGX Spark for the past few weeks. Not sure if anyone else is still optimizing this model on Spark, b…

@bernardlbmi3 @stefan132 @paxren2020 I have made a separate thread to discuss work around and optimizations of Qwen3.5-35B-A3B. Let’s continue at Qwen3.5-35B-A3B optimizations on single Spark

Your finding matches my real-world experience exactly. When using 122B INT4 for coding tasks, I noticed it would sometimes enter very long thinking loops and occasionally get stuck without producing any output — especially on complex multi-step problems. 35B Native FP8 just… gets it done. Faster th…

[image] chrm: I wonder, can we add the great 51 tok/s solution concept to your docker community image or it this solution already working as tested by myself (although I have not compared the mtp accuracy, but both are working for programming)? I don’t think any modifications for the docker b…

In brief, this thread pins vllm=0.19.0 because it depends on two different patches. vLLM’s architecture has changed since release and the patches - definitely the one to inc.py are not forward compatible right now. One of them is applied at build time for the hybrid FP8+int4 model. It is required f…

I think this patch is needed in your docker build to get extra tps boost: DGX_Spark_Qwen3.5-122B-A10B-AR-INT4/patches/03-int8-lm-head/patch_int8_lmhead.py at master · albond/DGX_Spark_Qwen3.5-122B-A10B-AR-INT4 · GitHub

I’m running aider bench agains original intel int4 autoround (finished), then this mod with merged layers and then I’ll try yours. Bench run takes ~3h so hopefully I’ll have initial results in ~7-8h

Model works and performance stayed the same. First tests also show good behaviour. Need more long time bench and work with it. But will use it for my daily work from now and report back :) thank you! Edit: At first glance it seems to hallucinate less and be a bit more accurate, lets see! It didnt …

awesome, thanks for that. will test it when I get home later today.

What is this reffering to? Albons build or something else?

Qwen3.5-122B-A10B on single Spark: up to 51 tok/s (v2.1 — patches + quick-start + benchmark)

Accelerated Computing DGX Spark / GB10 User Forum DGX Spark / GB10

grindstone April 13, 2026, 9:14am 246

Qwen3.5-122B-A10B-int4-AutoRound-EC Extended Calibration (EC)

https://huggingface.co/shieldstar/Qwen3.5-122B-A10B-int4-AutoRound-EC

Highly experimental, seems to pass initial testing, passed some comparison testing against both models. And i use it personally pretty hard.

All feedback is welcome, open to make more variants if it makes the ecosystem better.

@arctic.gus @norman.2

Best Q4 / NVFP4 model for quality Qwen3.5-27B or alternatives?

Topic		Replies	Views
Qwen/Qwen3.5-122B-A10B - Alibaba/Qwen thought about us... :-D DGX Spark / GB10	340	17836	March 24, 2026
Qwen/Qwen3.6-35B-A3B (and FP8) has landed DGX Spark / GB10 agentic-ai	309	30694	June 22, 2026
Qwen3.5-122B-A10B NVFP4 Quantized for DGX Spark — 234GB → 75GB, Runs on 128GB DGX Spark / GB10 Projects	44	12564	April 9, 2026
Does Qwen3.5-35B-A3B on GB10 leave a lot of performance on the table? DGX Spark / GB10 agentic-ai	40	6594	March 16, 2026
Qwen3.5-35B-A3B optimizations on single Spark DGX Spark / GB10 Projects	48	3759	May 22, 2026
What's the best speed we can get with Qwen 3.6 27B without quantizing? DGX Spark / GB10	64	23888	July 6, 2026
Qwen3.5-122B-A10B on single Spark: 15 → 21.5 tok/s with hybrid GPTQ-INT4 + FP8 dense layers (https://github.com/rmstxrx/vllm-hybrid-quant) DGX Spark / GB10 cuda	9	933	March 20, 2026
Qwen3.5-397B-A17B run in dual spark! but I have a concern DGX Spark / GB10	236	10110	June 6, 2026
Qwen3.5-397B-A17B + DGX Spark (duo) DGX Spark / GB10 Projects	62	6741	June 14, 2026
HOW-TO: Run Qwen3-Coder-Next on Spark DGX Spark / GB10 llama	92	11070	March 24, 2026

Qwen3.5-122B-A10B on single Spark: up to 51 tok/s (v2.1 — patches + quick-start + benchmark)

Related topics