Local agentic coding Benchmark: Qwen3.8-Flash-Next NVFP4 vs 27B (and the others...)
Heat trend
The percentage is based on available heat signal, not comment count or independent people.
Using https://huggingface.co/RadixArk/Qwen3.8-Flash-Next-NVFP4 and https://old.reddit.com/r/BlackwellPerformance/comments/1w04xb7/qwen38_flashnext_on_1x_rtx_pro_6000_171_ts_c1_428/
As usual, all the details in https://wonderrico.github.io/local_llm_benchmark/benchmark-main.html?filter=3.8 and even more in https://wonderrico.github.io/local_llm_benchmark/benchmark-detail.html?filter=3.8
(the bad score one is a "random" uncensored version from HF https://huggingface.co/dealignai/Qwen3.8-Flash-Next-UNCENSORED-NVFP4) I shall test other ones
Bottom line: almost highest score of all local model I tested, the most efficient in both nb requests / point and fewer generated tokens / pt, all in medium reasoning. (xhigh is not useful, again, in this benchmark) and if it was not enough very fast
https://preview.redd.it/dnk0yc90g5mh1.png?width=1366&format=png&auto=webp&s=115509753cb30cfe67f9b9d13158dcc300c3435d