I had a chance to play with lukealonso/MiniMax-M2.7-NVFP4 and wanted to share my very first results (Dual Node Setup - 2x Asus Ascent GX10). I let vLLM calculate context and was able to get 196608 – not too terrible. Benchmarks: | model | test | t/s | peak t/s | …

This is for the nvfp4 one based on @ekkis with subtle tweaks: | model | test | t/s | peak t/s | ttfr (ms) | est_ppt (ms) | e2e_ttft (ms) | |:-------------|-----------------:|----------------:|-------------:|-------------------:|----------------…

Thanks for the update to the model, I was waiting for it, thanks for the new recipes

2x Asus Ascent GX10, performance very similar to M2.5 (which makes sense, basically same model, same size). model test t/s peak t/s ttfr (ms) est_ppt (ms) e2e_ttft (ms) cyankiwi/MiniMax-M2.7-AWQ-4bit pp2048 3121.55 ± 32.45 779.28 ± 6.82 656.16 ± 6.82 779.35 ± 6.82 cyankiwi/MiniMax-M2.7-A…

So if I am understanding the benchmarks right, it’s a little bit smarter than Qwen 3.5 397B, but also a little bit slower? You also lose out on image capabilities with Minimax 2.7. I’ve been super impressed with Qwen 3.5 and it’s been my daily driver. My only complaint is speed.

It’s far more reliable for tool calling – it also seems to be more consistent for agentic coding (I tested it with OpenClaw, Hermes, Pi, OpenCode). It doesn’t do vision and has less context. Ultimately you may want to test both and see which works better for you.

Any chances that his will run on a single Spark with REAP like: [image] MiniMax 2.5 REAP - NVFP4 on single DGX Spark DGX Spark / GB10 Yesterday a REAP version of MiniMax 2.5 showed up already quantised to NVFP4: ran benchy on it: model test t/s peak t/s …

MiniMax family less AGI and more coder/tests passer than Qwen, which is all rounder.

[image] cosinus: More important for us in here I think: https://x.com/RyanLeeMiniMax/status/2043580596245700906 …so as long as you pile up your Sparks at home and use it, you should be fine. 😅 That sounds good, will download MiniMax M2.7 then :)

I can only speak to the full FP8 weights at the moment, but I also confirm that from Qwen 397B FP8 to MiniMax M2.7 FP8 that M2.7 is way, way better at writing test coverage. Using the same script, Qwen only produced about 30 tests on its own and it took a massive number of iterations (like a couple …

Give Qwen3-Coder-Next a spin. It’s very, very fast and if you don’t care about image capabilities I still think it’s the best coder and agentic model out there. The NVFP4 version is still very capable even with the quantization.

MiniMax M2.7 NFVP4 Recipe & Benchmarks

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

carlos.albarran.mx April 13, 2026, 2:59pm 23

Any chances that his will run on a single Spark with REAP like:

Topic		Replies	Views
MiniMax M2.5 released (not available on HuggingFace as of now) -- is DGX Spark ready? DGX Spark / GB10	92	6154	April 12, 2026
MiniMax 2.5 REAP - NVFP4 on single DGX Spark DGX Spark / GB10	25	2863	April 1, 2026
Can someone with 2 Sparks benchmark NVFP4 MiniMax M2.1 quant? DGX Spark / GB10	25	1428	January 29, 2026
MiniMax-2.5 on DGX Spark (thanks to Unsloth https://unsloth.ai/docs/models/minimax-2.5) DGX Spark / GB10 llama	12	3636	February 20, 2026
Help: Running NVFP4 model on 2x DGX Spark with vLLM + Ray (multi-node) DGX Spark / GB10 mistral-large	18	2393	December 25, 2025
DGX Spark performance DGX Spark / GB10	50	4573	February 27, 2026
Qwen3.6-27B is out! DGX Spark / GB10 agentic-ai	54	9338	May 4, 2026
RedHatAI/Qwen3.5-122B-A10B-NVFP4 seems to be the best option for a single Spark DGX Spark / GB10 Projects llm	74	5053	April 11, 2026
Two-Spark cluster with vLLM using tensor-parallel-size 2 causes one node to drop while the other's GPU goes 100% forever DGX Spark / GB10	36	1339	February 13, 2026
We unlocked NVFP4 on the DGX Spark: 20% faster than AWQ! DGX Spark / GB10	145	7157	March 28, 2026

MiniMax M2.7 NFVP4 Recipe & Benchmarks

Related topics