conv.

All stories
Active · 25h ↳ Antirez/h3.c: MiniMax H3 inference engine for Mac computers

Antirez/h3.c: MiniMax H3 inference engine for Mac computers

14 items427 engagement94 comments Hacker NewsPressMastodon
Antirez/h3.c: MiniMax H3 inference engine for Mac computers
github.com

Conversation activity · last 26 hours peak 3/30m

Peak 3 items in one 30m at Aug 10, 8 PM; 14 items over 26 hours Aug 10, 8:25 PM — no itemsAug 10, 8:55 PM — 3 items · Press 2, Hacker News 1Aug 10, 9:25 PM — no itemsAug 10, 9:55 PM — no itemsAug 10, 10:25 PM — 2 items · Hacker News 1, Mastodon 1Aug 10, 10:55 PM — 1 item · Hacker News 1Aug 10, 11:25 PM — no itemsAug 10, 11:55 PM — 1 item · Hacker News 1Aug 11, 12:25 AM — no itemsAug 11, 12:55 AM — no itemsAug 11, 1:25 AM — no itemsAug 11, 1:55 AM — 1 item · Hacker News 1Aug 11, 2:25 AM — 1 item · Hacker News 1Aug 11, 2:55 AM — no itemsAug 11, 3:25 AM — no itemsAug 11, 3:55 AM — 1 item · Hacker News 1Aug 11, 4:25 AM — 1 item · Hacker News 1Aug 11, 4:55 AM — no itemsAug 11, 5:25 AM — no itemsAug 11, 5:55 AM — no itemsAug 11, 6:25 AM — no itemsAug 11, 6:55 AM — 2 items · Hacker News 2Aug 11, 7:25 AM — no itemsAug 11, 7:55 AM — no itemsAug 11, 8:25 AM — no itemsAug 11, 8:55 AM — no itemsAug 11, 9:25 AM — no itemsAug 11, 9:55 AM — no itemsAug 11, 10:25 AM — no itemsAug 11, 10:55 AM — no itemsAug 11, 11:25 AM — no itemsAug 11, 11:55 AM — no itemsAug 11, 12:25 PM — no itemsAug 11, 12:55 PM — no itemsAug 11, 1:25 PM — no itemsAug 11, 1:55 PM — no itemsAug 11, 2:25 PM — no itemsAug 11, 2:55 PM — no itemsAug 11, 3:25 PM — no itemsAug 11, 3:55 PM — no itemsAug 11, 4:25 PM — no itemsAug 11, 4:55 PM — 1 item · Mastodon 1Aug 11, 5:25 PM — no itemsAug 11, 5:55 PM — no itemsAug 11, 6:25 PM — no itemsAug 11, 6:55 PM — no itemsAug 11, 7:25 PM — no itemsAug 11, 7:55 PM — no itemsAug 11, 8:25 PM — no itemsAug 11, 8:55 PM — no itemsAug 11, 9:25 PM — no itemsAug 11, 9:55 PM — no items 3 items · 8:55 PM
Aug 118 AM4 PMnow · 10:25 PM

Clustered from 14 items across 3 sources. Not yet parsed — the coverage below is the raw record.

Press coverage 2

Social posts 2

Voices from the web unedited

  • I've been using MiniMax H3 on my M5 Pro 64GB MacBook Pro through ComfyUI. It works extremely well.I had to modify the default ComfyUI workflows to use a GGUF quant (city96's ComfyUI-GGUF custom node, UnetLoaderGGUF in place of the stock loader) [0].I use the model labeled Q5_K_M. There is Q8_0 available as well, which is 34GB and fits fine in 64GB…

    MeleagrisHacker News23h agoview on Hacker News ↗

  • GGUF is outdated in the latest versions of Comfy-UI. If you want a good balance of size, speed and quality you should use the int8_convrot model from the official Comfy Org Repo

    vimtoHacker News15h agoview on Hacker News ↗

  • > a ~9-second 480x864 clip at 20 steps takes me a bit over an hourthat's rough. for comparison, i tried the exact same parameters on my 5090 RTX and it took 2 minutes to generate.i believe diffusion models are primarily compute bound so the macs aren't really the ideal hardware for this kind of stuff

    thousand_nightsHacker News15h agoview on Hacker News ↗

  • In the AMA Minimax said that H3 could support sparse attention, that would be a huge speedup! I wonder if there are any news on that. H3 is very cool. EDIT: testing a --sparse-attention optional mode based on what they said in the Reddit post.

    antirezHacker News20h agoview on Hacker News ↗

  • On my 128GB M4 Max Mac Studio, generating a 15s 480p video with MiniMax H3 in ComfyUI takes an hour and a half.Put Codex to work on deploying it now, hoping the speed can improve quite a lot :-) Thanks anyway

    linzhangrunHacker News19h agoview on Hacker News ↗

  • This implementation is much faster on my M5 Max, like a few minutes for the same video, but on an M5 Max with 128GB, didn't test on M5 Pro. About memory, could be executed on 64GB with a few changes.

    antirezHacker News18h agoview on Hacker News ↗

  • This is where the DGX spark makes up a bit of the ground it loses on llm work, diffusion and cuda go together like peanut butter and jelly.

    diddidHacker News22h agoview on Hacker News ↗

  • Native inference optimization for Apple Silicon is such a game-changer for local-first workflows. Incredible performance work.

    myshapeprotocolHacker News17h agoview on Hacker News ↗

  • This still requires 128Gb of memory, right? Me and my lowly 96Gb, like a commoner; missing out on the fun.

    TechSquidTVHacker News23h agoview on Hacker News ↗

More in this narrative