FraShin

🇮🇹 Como Lake
AI Shaping Reality / Tech & Systems
Single-Player Immersion / Fudōshin (不動心)
Socials + Live on Twitch 👇

[FraShin @FL511_]

This is a masterclass in storytelling, probably the best piece of speculative writing on this app. Exaggerated? Maybe. But it grips you like a psychological thriller. If it’s fiction, it’s top-tier. If it’s not... well 👀

[🍓🍓🍓 @iruletheworldmo]

https://t.co/GCz7g67znl

[FraShin @FL511_]

@coldniko No way! Or? 👀

[FraShin @FL511_]

@AnkitKumar65837 @coldniko You also need 64 gb ram

[FraShin @FL511_]

@JustLingonberry Huge if true

[FraShin @FL511_]

@alexandr_wang Cool but Europe when?

[FraShin @FL511_]

@laurent_zw are you getting better result with this model compared to the ista one?

[FraShin @FL511_]

@seanhighness @Eschalabs good job, dropping when?

[FraShin @FL511_]

@sudoingX ptq1 on my 4080 it goes from ~47 tok/s to ~59 tok/s with 262k ctx and 14.2/15.2 gb vram

[FraShin @FL511_]

@sudoingX 👀

[FraShin @FL511_]

@AnfieldPapers lmaooo

[FraShin @FL511_]

@Anfield_Sector Warming the bench as usual.

[FraShin @FL511_]

@BenBocsak In the world.

[FraShin @FL511_]

@Anfield_Sector Wirtz off for the love of this sport. He’s terrible

[FraShin @FL511_]

@TaintlessRed He’s a problem, a big one

[FraShin @FL511_]

@ekitikEraa He’s dogshit, it wouldn’t be that hard to do better

[FraShin @FL511_]

@AnfieldEdition Thanks kerkez, wirtz should he off already

[FraShin @FL511_]

@Lfc_samie He’s so fucking dogshit man

[FraShin @FL511_]

@Anfield_Sector Wirtz is dogshit I can’t stand him anymore

[FraShin @FL511_]

@sudoingX Yep, this model is amazing..I don’t understand the hate 🤷🏻‍♂️🤷🏻‍♂️

[FraShin @FL511_]

WOW! Tested Bonsai on Hermes Agent and I'm in awe. Running PTQ1 on a 4080 w/ 262k context 14,9 gb vram ~59 tok/s decode, ~1k prefill.
Crushed 4 hard tests: logic, code, vision and agentic tool loops (autodebug, scraping) on medium reasoning.
Incredible engineering @PrismML 🚀🚀

[PrismML @PrismML]

Today, we’re announcing Ternary Bonsai 2 27B.

Based on Qwen3.8 27B, Bonsai 2 27B is 9x smaller than its full-precision counterpart while retaining 98.2% of its aggregate benchmark performance.

Two months after the first Bonsai 27B release, the biggest change is quality. The footprint remains 5.9 GB, but the gap to full precision has narrowed materially, with particularly strong gains in agentic coding, multimodal reasoning, and long-horizon tool use.

Ternary Bonsai 2 27B is available today under Apache 2.0.

[FraShin @FL511_]

@NVIDIAGeForce #RTXPowersPlay 🫡🫡