FRAMEWIREIndonesiaUpdated Sep 14Live wire
0:00 / 0:00

It's because Voice mode doesn't use reasoning, see my local qwen 3.8 27b example

First answer with no reasoning, second answer with reasoning set to low

ErwinSep 1315
0:00 / 0:00

Qwen 3.8 Next Flash 4bit MTP on my M3U Studio is absolutely flying now.

My latest OMP run measured 97.1 tok/s overall decode at 23k context, with ~1,131 tok/s uncached PP. PR opened on oMLX for this.

Ash HartSep 1377
0:00 / 0:00

Holy smokes qwen-3.8-27b on cerebras with Hermes Agent feels so good and fast it's almost unreal.

Testing with the upcoming Cadu app by fl_rn_st. Tool calls are now the main slowdown 😅

Federico ViticciSep 13143
0:00 / 0:00

DeepSeek V4.1 Flash is now listed with free access on Apinex—and the catalog includes several other models

DeepSeek V4 Pro • DeepSeek V4 Flash • Gemini 3.8 Flash • GLM 5.3 Flash • Muse Spark 1.3 • GPT-5.6 Luna • Qwen 3.8 Max Browse the models:

Tung AirSep 13