Trying out now with 5bit quant and 1M context !

There is a video now with 1M Context youtu.be/TMFCDb6mGBY

There is a video now with 1M Context youtu.be/TMFCDb6mGBY

Running QWEN 3.8-FLASH-NEXT with 1Million Context with Strata in Apple M5 Max @coldniko youtu.be/TMFCDb6mGBY?si=1ZLC… #executeautomation #strata

@liquidai LFM2.5-24B Hits 70 TPS on @Apple Mac mini M6, I am executing the model for OpenClaw with 128K Context! Checkout here 👇 youtu.be/cSvY35P434o #executeautomation #liquidai

@liquidai LFM2.5-24B Hits 70 TPS on @Apple Mac mini M6, I am executing the model for OpenClaw with 128K Context! Checkout here 👇 youtu.be/cSvY35P434o #executeautomation #liquidai

Im going to do a run today and keep you all posted !

Tried with 1M context length in my Apple M5 Max Got 131 TPS with just 70GB of RAM used

1Million Token Context for QWEN3.8-Flash-Next Model in Strata of @coldniko Getting 131 TPS in Apple M5 Max Tell me if I am doing something wrong or Strata is good 😊

@trymirai Uzu on Apple M5 max and Apple M6 Mini We are getting Faaaaast Tokens in local machine 💨 youtu.be/0xNgYkvWVSQ?si=4zjQ…

@coldniko - see how your work is inspired here ❤️

Great work bro and yes I love the speed. Waiting for more improvements and models support from you !

That’s awesome bro ! Will try it out now

Unfortunately, yes, bro—please release the Apple Silicon version. I will showcase it and explain how Strata will render other inference engines as Stone Age 🤣

Bro what is the inferencing engine and where is it running and config of the machine if its local ?

Both please ! Small - QWEN4 - 27B, QWEN4-45B-A6B (for a change) Large - QWEN4-Flash-Next 125B

Respect sir 🫡

@grok when is the release date of QWEN-4 27b ?