Conversation

Notices

1. Embed this notice
   cell classic (cell@shitposter.world)'s status on Friday, 24-Jul-2026 19:00:29 JST cell classic
   all the latest and greatest LLMs, be it proprietary SaaS or open-weight focus on complex tasks like programming, math etc.
   
   but for basic writing and document handling, do they still require such large models, or is a lightweight model that can run on your laptop or phone sufficient?
   • Embed this notice
     lainy (lain@lain.com)'s status on Friday, 24-Jul-2026 19:00:28 JST lainy

     in reply to
     @cell i don't think so. gemma 4 and qwen should be fine for that.
   • Embed this notice
     lainy (lain@lain.com)'s status on Friday, 24-Jul-2026 19:04:08 JST lainy

     in reply to
     @cell yeah, frontier models are now actually approaching the level of 'make gta 6, make no mistakes', so basic agentic tasks don't get much news anymore
   • Embed this notice
     cell classic (cell@shitposter.world)'s status on Friday, 24-Jul-2026 19:04:09 JST cell classic

     in reply to

     • lainy
     @lain i mean i've been using those and the basic models from big providers (e.g. claude sonnet, gemini flash, gpt instant) for most of those "basic" tasks already :cat_si:
     
     seems like a solved problem already if you're not architecting anything complex so i don't think alot of attention is directed at this sector anymore :cat_eyeonyou:
   • Embed this notice
     ᴏᴏᴍ-ᴋɪʟʟᴇʀ: 333[unix.exe]™ (jae@mastodon.bsd.cafe)'s status on Friday, 24-Jul-2026 19:06:03 JST ᴏᴏᴍ-ᴋɪʟʟᴇʀ: 333[unix.exe]™

     in reply to

     • lainy

     @lain @cell also some of the smaller moe models or 1bits could lift this.

     lainy likes this.
   • Embed this notice
     Blurry Moon (sun@shitposter.world)'s status on Friday, 24-Jul-2026 19:27:03 JST Blurry Moon

     in reply to
     @cell I used deepseek4 effectively