For those of you who don’t want to use a ChatGPT but want a LLM.

  • sir_reginald@lemmy.world
    link
    fedilink
    English
    arrow-up
    19
    ·
    edit-2
    8 months ago

    As it’s to be expected from this website, the article quality is subpar.

    They mix third party hosted LLMs, with frontends (KoboldAI, GPT4ALL) and with local models (which BTW, every single one of them is a llama finetune. They’ve never heard of mistral or others, apparently).

    If I hadn’t read this website years ago and knew that the quality was already trash back then, I’d say this very article was written by a LLM.

  • BetaDoggo_@lemmy.world
    link
    fedilink
    English
    arrow-up
    11
    ·
    8 months ago

    Very strange article. It lists several front ends, some of which are not open source, as well as some raw models without clearly distinguishing between them. RWKV was mentioned which is cool I guess.

    The first option listed should have been huggingchat, followed by the various local UIs, with a separate section discussing the models themselves.

    • Chewy@discuss.tchncs.de
      link
      fedilink
      English
      arrow-up
      5
      ·
      8 months ago

      I feel like most of these “10 alternatives to xyz”-articles are basically a summary of alternativeto.net. Or they’ve just listed all projects they’ve found with a quick search. I’m almost certain they didn’t install them most of the time.

      This also applies to “comparison” sites, which usually are a list of Amazon affiliate links. At this point, I don’t trust websites with affiliate links anymore, as they’ve never actually tried the products. Sadly those spam sites make it difficult to find actually good reasearched tests.

      Back to itsfoss, they write many articles, and some are good, but they still are blog spam.

  • cyd@lemmy.world
    link
    fedilink
    English
    arrow-up
    8
    ·
    edit-2
    8 months ago

    Pretty bad article, but here are my two cents on the actual answer, as of right now:

    • General chatting: OpenHermes
    • General querying, instruction: Mistral
    • Coding: Deepseek-coder

    These perform similar to or better than ChatGPT 3.5, in some cases comparable to 4.

    For specific niche applications (role playing, nsfw stories, etc), just search on huggingface.co and read the user reviews.

  • General_Effort@lemmy.world
    link
    fedilink
    English
    arrow-up
    6
    arrow-down
    1
    ·
    8 months ago

    It’s a bad list.

    YSK: There are several things mixed in it.

    1. There are “models” (usually meaning “weights” specifically). The models (weights) are the raw data that contain the wisdom of the AIs.
    2. There is software to “run” these models on your own machine, or to connect to an API.
    3. There are services which run models for you and let you interact via web interface, app, or API. Some services may add text to your prompt, to create a better(?) prompt.

    There is a bewildering array of models out there. Mostly, they are specialized and/or merged versions of some popular foundation models (by mistral, meta, and a few others). Without endorsing any service, I find that openrouter.ai and together.ai let you try a fairly large selection. There are other services.

    You can find more/better information here: !localllama@sh.itjust.works

  • Substance_P@lemmy.world
    link
    fedilink
    English
    arrow-up
    1
    ·
    edit-2
    8 months ago

    I’ve been using iask.ai or Phind.com for mostly quick answers to my simple questions, these get me by but I’m no coder or author and I have noticed these options are often pretty terrible for math queries.

    Edit: the above services I’m pretty sure are not open source, but their privacy policies seem reasonable.