Kaia@kaia@pleroma.soykaf.com: @lain very jealous :peek_cat:

Home | Notifications | New Note | Local | Federated | Search | Logout

Note Detail

lain@lain@fediffusion.art (2026-09-09 17:01:27)
both qwen 3.8 flash next (roughly opus 4.6 level, i.e 'actually good) and krea 2 fit into the strix halo and run at good speed for agentic use. so it took only around 9 months from good hosted LLMs to good local LLMs.
---Reply---
Kaia@kaia@pleroma.soykaf.com (2026-09-09 17:07:20)
@lain very jealous :peek_cat:


---Replies---

lain@lain@fediffusion.art (2026-09-09 17:11:24)
@kaia i think these unified memory things will become more popular, now that there's actually a reason to have all that vram. and maybe the chinese will come out with a chip that does 70% speed of 4090 but has 64gb ram, that would shake things up as well