Qwen3.6-27B is out ! i hope for 122B now [image] Qwen/Qwen3.6-27B · Hugging Face We’re on a journey to advance and democratize artificial intelligence through open source and open science.

Dual Node with MTP. I am still testing this but it seems like each speculative token adds an allgather across the inter-node link. With num_speculative_tokens=2 , that’s 2 extra cross-node round trips per decode step on top of the normal allreduce — likely 2-3× the communication overhead, eating any…

I ran the same subset of AgentBench with the FP8 version, and amazingly it was faster than the MoE (also FP8) version. I can only assume it went in less circles or generated less errors calling tools. [Screenshot 2026-04-22 at 23.32.14] (I also can’t explain why the FP8 MoE beat the bf16 MoE, but…

I’m thinking that, although there are improvements over 3.5, overall the model is just thinking too much and getting bogged down without the prospect of an actual result. Testing with cyankiwi/Qwen3.6-27B-AWQ-INT4 I get decent responses from fairly simple prompts, but when I throw it at a real-worl…

Which coding agent do you use? Claude Code? OpenCode? VS Code CoPilot?

[image] DannyTup: I’ll try to kick off the bf16 version soon. I posted the updated results here: [image] Collecting eval results for Spark-sized quants of models DGX Spark / GB10 I ran the same subset of AgentBench (3 times per sample, taking the mean r…

I just tried to run a similar llama-bency as posted by @serapis , for both Qwen3.6-35B-A3B and -27B. Running llama-cpp b8672 on 2 NV-linked RTX3090. Using Unsloth’s UD-Q4_K_XL quantizations for both models. Qwen3.6-27B llama-benchy --base-url "http://127.0.0.1:8080/v1" --model "Qwen/Qwen3.6-27B"…

Just saw this post @iotcoi on X: Qwen3.6-27B-FP8 + Dflash + DDTree, 256k context, 10 agents ~200 tokens/sec max decode 136t/s average on a single tiny GB10 GPU at 49W power I will not post the link so I am not flagged here, but he was the first to make DDTree work on spark but, looks like, is not …

yep, DDTree looks very promising, just need someone with better technical skillz than mine to bolt it into vllm.

[image] danielfl: @iotcoi He said he will clean it up and put it here: [image] GitHub - mitkox/vllm-dflash-ddtree: vLLM experimental implementation of DDTree with... vLLM experimental implementation of DDTree with Dflash It is empty now, but h…

Qwen3.6-27B is out!

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

DannyTup April 23, 2026, 7:09am 15

I posted the updated results here:

Topic		Replies	Views
Qwen/Qwen3.6-35B-A3B (and FP8) has landed DGX Spark / GB10 agentic-ai	309	31198	June 22, 2026
What's the best speed we can get with Qwen 3.6 27B without quantizing? DGX Spark / GB10	64	24723	July 6, 2026
Qwen3.5 27B optimisation thread starting at 30+ t/s TP=1 DGX Spark / GB10 llama , agentic-ai	23	3217	May 11, 2026
Qwen3.6-27B-Dflash link DGX Spark / GB10 Projects	20	5554	April 29, 2026
Benchmark Report: Qwen/Qwen3.6-27B vs nvidia/Qwen3.6-27B-NVFP4 DGX Spark / GB10 Projects	5	2601	July 4, 2026
Fastest Qwen 3.5 122B Int4 recipe on DGX Spark tested and published on Spark-Arena DGX Spark / GB10 llama	59	4758	June 3, 2026
New 2.5x Faster Qwen3.6 NVFP4 Unsloth quants DGX Spark / GB10	20	4150	July 13, 2026
Bfloat16 Quality = Speed? DGX Spark / GB10	106	6933	May 26, 2026
DFlash for Qwen3.5-122B-A10B = 80+ tok/s on 1x Spark! DGX Spark / GB10 Projects agentic-ai	72	9293	July 27, 2026
Single-Spark setups — which models do you actually run for coding, and how? (sharing mine + a test prompt) DGX Spark / GB10 spark , llama , agentic-ai , deepseek , dgx	59	7177	August 6, 2026

Qwen3.6-27B is out!

Related topics