← Back to context

Comment by johnmaguire

1 day ago

I'd love to know how you fit smaller models into your workflow. I have an M4 Macbook Pro w/ 128GB RAM and while I have toyed with some models via ollama, I haven't really found a nice workflow for them yet.

32 comments

johnmaguire

Reply
philipkglass  1 day ago

It really depends on the tasks you have to perform. I am using specialized OCR models running locally to extract page layout information and text from scanned legal documents. The quality isn't perfect, but it is really good compared to desktop/server OCR software that I formerly used that cost hundreds or thousands of dollars for a license. If you have similar needs and the time to try just one model, start with GLM-OCR.

If you want a general knowledge model for answering questions or a coding agent, nothing you can run on your MacBook will come close to the frontier models. It's going to be frustrating if you try to use local models that way. But there are a lot of useful applications for local-sized models when it comes to interpreting and transforming unstructured data.

  • mandeepj  1 day ago

    > I formerly used that cost hundreds or thousands of dollars for a license

    Azure Doc Intelligence charges $1.50 for 1000 pages. Was that an annual/recurring license?

    Would you mind sharing your OCR model? I'm using Azure for now, as I want to focus on building the functionality first, but would later opt for a local model.

    • philipkglass  21 hours ago

      I took a long break from document processing after working on it heavily 20 years ago. The tools I used before were ABBYY FineReader and PrimeOCR. I haven't tried any of the commercial cloud based solutions. I'm currently using GLM-OCR, Chandra OCR, and Apple's LiveText in conjunction with each other (plus custom code for glue functionality and downstream processing).

      Try just GLM-OCR if you want to get started quickly. It has good layout recognition quality, good text recognition quality, and they actually tested it on Apple Silicon laptops. It works easily out-of-the-box without the yak shaving I encountered with some other models. Chandra is even more accurate on text but its layout bounding boxes are worse and it runs very slowly unless you can set up batched inference with vLLM on CUDA. (I tried to get batching to run with vllm-mlx so it could work entirely on macOS, but a day spent shaving the yak with Claude Opus's help went nowhere.)

      If you just want to transcribe documents, you can also try end-to-end models like olmOCR 2. I need pipeline models that expose inner details of document layout because I need to segment and restructure page contents for further processing. The end-to-end models just "magically" turn page scans into complete Markdown or HTML documents, which is more convenient for some uses but not mine.

      3 replies →

tempaccount5050  20 hours ago

Not OP but I had an XML file with inconsistent formatting for album releases. I wanted to extract YouTube links from it, but the formatting was different from album to album. Nothing you could regex or filter manually. I shoved it all into a DB, looked up the album, then gave the xml to a local LLM and said "give me the song/YouTube pairs from this DB entry". Worked like a charm.

Bluecobra  1 day ago

I didn’t realize that you can get 128GB of memory in a notebook, that is impressive!

  • lambda  1 day ago

    I've got a 128 GiB unified memory Ryzen Ai Max+ 395 (aka Strix Halo) laptop.

    Trying to run LLM models somehow makes 128 GiB of memory feel incredibly tight. I'm frequently getting OOMs when I'm running models that are pushing the limits of what this can fit, I need to leave more memory free for system memory than I was expecting. I was expecting to be able to run models of up to ~100 GiB quantized, leaving 28 GiB for system memory, but it turns out I need to leave more room for context and overhead. ~80 GiB quantized seems like a better max limit when trying not running on a headless system so I'm running a desktop environment, browser, IDE, compilers, etc in addition to the model.

    And memory bandwidth limitations for running the models is real! 10B active parameters at 4-6 bit quants feels usable but slow, much more than that and it really starts to feel sluggish.

    So this can fit models like Qwen3.5-122B-A10B but it's not the speediest and I had to use a smaller quant than expected. Qwen3-Coder-Next (80B/3B active) feels quite on speed, though not quite as smart. Still trying out models, Nemotron-3-Super-120B-A12B just came out, but looks like it'll be a bit slower than Qwen3.5 while not offering up any more performance, though I do really like that they have been transparent in releasing most of its training data.

    • zozbot234  1 day ago

      There's been some very recent ongoing work in some local AI frameworks on enabling mmap by default, which can potentially obviate some RAM-driven limitations especially for sparse MoE models. Running with mmap and too little RAM will then still come with severe slowdowns since read-only model parameters will have to be shuttled in from storage as they're needed, but for hardware with fast enough storage and especially for models that "almost" fit in the RAM filesystem cache, this can be a huge unblock at negligible cost. Especially if it potentially enables further unblocks via adding extra swap for K-V cache and long context.

  • AzN1337c0d3r  1 day ago

    Most workstation class laptops (i.e. Lenovo P-series, Dell Precision) have 4 DIMM slots and you can get them with 256 GB (at least, before the current RAM shortages).

    There's also the Ryzen AI Max+ 395 that has 128GB unified in laptop form factor.

    Only Apple has the unique dynamic allocation though.

    • the_pwner224  1 day ago

      Yep, I have a 13" gaming tablet with the 128 GB AMD Strix Halo chip (Ryzen AI Max+ 395, what a name). Asus ROG Flow Z13. It's a beast; the performance is totally disproportionate to its size & form factor.

      I'm not sure what exactly you're referring to with "Only Apple has the unique dynamic allocation though." On Strix Halo you set the fixed VRAM size to 512 MB in the BIOS, and you set a few Linux kernel params that enable dynamic allocation to whatever limit you want (I'm using 110 GB max at the moment). LLMs can use up to that much when loaded, but it's shared fully dynamically with regular RAM and is instantly available for regular system use when you unload the LLM.

      7 replies →

    • lambda  1 day ago

      > Only Apple has the unique dynamic allocation though.

      What do you mean? On Linux I can dynamically allocate memory between CPU and GPU. Just have to set a few kernel parameters to set the max allowable allocation to the GPU, and set the BIOS to the minimum amount of dedicated graphics memory.

      3 replies →

    • numpad0  15 hours ago

      Intel had dynamic allocation since Intel 830(2001) for Pentium III Mobile. Everything always did, especially platforms with iGPUs like Xbox 360.

      Only Apple and AMD have APUs with relatively fast iGPU that becomes relevant in large local LLM(>7b) use cases.

saltwounds  1 day ago

I use Raycast and connect it to LM Studio to run text clean up and summaries often. The models are small enough I keep them in memory more often than not

echelon  1 day ago

Shouldn't we prioritize large scale open weights and open source cloud infra?

An OpenRunPod with decent usage might encourage more non-leading labs to dump foundation models into the commons. We just need infra to run it. Distilling them down to desktop is a fool's errand. They're meant to run on DC compute.

I'm fine with running everything in the cloud as long as we own the software infra and the weights.

This is conceivably the only way we could catch up to Claude Code is to have the Chinese start releasing their best coding models and for them to get significant traction with companies calling out to hosted versions. Otherwise, we're going to be stuck in a take off scenario with no bridge.