deepblue

Old

[deepblue @DeepBlue9696]

lol i broke qwen3.8 iq3xss on strata with this request on the webui: "write me a detailed report on Qwen ai status and architectures" ofc no tools so it keept thinking about it's cutoff date and somehow always ended in infinite looping xD

[deepblue @DeepBlue9696]

@ProtonVPN E, would rate the uk in the top 10 no digital privacy for citizens ranks

[deepblue @DeepBlue9696]

@ScottyT72 @coldniko my bad forgot that was the reason why i responded: context 262144

[deepblue @DeepBlue9696]

@ScottyT72 @coldniko I run on a 3090 + 64gb ddr4 1866mhz at 60tks and 1500-1800tks promptprocessing

[deepblue @DeepBlue9696]

@Birch_ONLINE @Vinci_x3 let me add: Custom Robo Arena and Spectrobes

[deepblue @DeepBlue9696]

@Birch_ONLINE @Vinci_x3 bro knows the greatness of some niche nds games😎

[deepblue @DeepBlue9696]

@ShackleOhNo your team was just weak or they were the ones searching for the speedup tool, I doubt you would even have time to stack the meter in a decent group...

[deepblue @DeepBlue9696]

I've called @coldniko project bs, then i tryied it and had to make it do a website just for it https://t.co/x7az95YnSY
3090 +64gb ddr4 1866mhz pp 1300-1800 gen 50-60 qwen falsh iq3xss

[deepblue @DeepBlue9696]

@steeve @Qualcomm @zml_ai llama 3.2 1b... i call it useless it's barely at the level of a phone

[deepblue @DeepBlue9696]

@ItsmeAjayKV hell no, 8k price + extremely low mem bandwidth is such a bad deal when we have the m5 and the xiaomi box coming soon

[deepblue @DeepBlue9696]

@0xSero no recipes for amd gpus lol still interesting choice the exllama format, unfortunately i'm not convinced that it's quality is at least on par of q6kl from unsloth.

[deepblue @DeepBlue9696]

@LottoLabs ppl are teasing this for 3years now. We have grown tired of such bs

[deepblue @DeepBlue9696]

@1337hero we still waiting the github repo... btw for tts did u use omnivoice?

[deepblue @DeepBlue9696]

@mfranz_on let me fix your comment: "almost" Good performance, "many" setup issues, supports almost every "common" model

[deepblue @DeepBlue9696]

@ItsmeAjayKV try having it do character in T-pose for usage with trellis i still wasn't able to have it properly comply. ps do you use comfy or something else for it?

[deepblue @DeepBlue9696]

@ItsmeAjayKV @PrismML like sh*t

[deepblue @DeepBlue9696]

@ItsmeAjayKV @unionalphaai this is trash but ok... qwen3.8 all versions (27b included) did a much better job than this stealth weakling

[deepblue @DeepBlue9696]

@unionalphaai why is everyone so hyped? the quality is mistral 24b level (trash)

[deepblue @DeepBlue9696]

@bet3rd @killa @Steam Useless, its standard admin installation i would work with any software that has a bg service even adobe suite