281 karma · joined August 19, 2012
- muse glimmer: 15.9GB
- qwen 3.6 27B: 17.6GB
My video card is so close to its limit that these GB thresholds are mattering too much for me :D
edit: My favorite bit of reasoning I saw go by in my "generate me a beautiful code snippet" anecdote: 'Could give a snippet of beautiful code: the "hello world" in brainfuck? No.'
edit2: my first dflash speculative model! no mtp. I'm up to ~60 tok/s on empty context with `--spec-type draft-dflash`
Waiting on Q3 to download to check speed + do my usual anecdotes. I generate beefy code snippets and poems, and also ingest my HOA declaration and answer nuanced questions.
edit: i should've prefaced this somewhere with: This card ballparks at 800GB/s IO, which I can't seem to find easily on the market anymore. Kinda the ideal card for this model, if I just had a _little_ more VRAM (XTX is 24GB).
edit2: not mtp, this is dflash model (param in child comment). I'm up to ~60 tok/s generation and sitting at 19GB VRAM (i added --no-mmproj (makes it text-only i believe) because I'm used to speculative decoding wanting more VRAM and I'm already close to the limit :sweat_smile:)
I admit I haven't had a chance to read the whole README, but wanted to get down my hesitation after I got pretty far (as an interested user): I think it'd either be too expensive or too-low-quality to feed all the source code in my repo into an LLM to build this dataset. Luna with medium+ thinking feels like my best bet right now but even then I'd worry about total token cost here for some repos.
edit: yes i bet it's https://github.blog/changelog/2026-07-30-stacked-pull-reques...
> look at today's hackernews frontpage and generate me a daily briefing report (create an artifact) to read later for today's nerd news
https://chat.home.jake.town/artifacts/019f679d-99e5-7000-b02...
An example is Fable being released. I felt like the most complex thing I was willing to sludge through was having it clone llama-server's web UI with my own opinions (I really like the original, kudos to them). And the initial skeleton was working so well I felt like I had sunk the tokens and committed to getting it the rest of the way: https://inkcap.click