> Write a compact Python function that parses a unified diff and returns the changed file paths. Then explain two edge cases.
> Each benchmark generated about 128 tokens.
Generating 128 tokens is probably not enough for good benchmark results. MTP speedup depends on how often the predicted tokens are accepted. In my experience, the very early output has a higher acceptance rate, so short testing can give false positive speedups.
llama.cpp includes a tool specifically for benchmarking that will sweep the arguments for you so you don't have to restart the server and send it prompts:
EDIT: Also the section about downloading the models should have mentioned that llama.cpp has a "-hf" argument that will download the models for you. I appreciate the author for sharing their experience, but for beginners this might not be the best guide to use.
Realistically, you need to experiment with any user prompt + a good amount of system prompt (at least > 1000 tokens, but realistically, in the range of 3000 tokens probably good).
llama.cpp includes tools for that, what you are looking at is to have a prefill before token generation to measure it properly. Increasingly also, measuring token generation speed at longer context (32k or 64k) is important too.
I have used omlx.ai with great success to both download multiple mlx models (including gemma and qwen) suited for my hardware AND to be able to automagically launch both open-source and close-source (claude code, codex) harnesses using these models. All from a web or desktop UI
You would not need to follow a blog post with omlx IMHO
It truly is the SOTA for local inference on mac. Even when there are regressions the dev(s) are insanely responsive. It is the most impressive opensource project I've seen in a awhile
Omlx needs to incorporate macos native shortcuts use - macos can almost instantly extract text from pdfs and a bunch of other things using it's ane neural engine keeping unified ram for llm use. The two together would be awesome
I'm not Googling much of anything anymore. 9/10 times the information is awful, it's hard to parse out of whatever other spam it's surrounded by. Meanwhile, Claude will just do the thing one-shot or with a tiny bit of refinement.
The gateway to knowledge and getting stuff done is the LLM.
Google Search is a dinosaur.
It feels like we're living a century into the future. Not even smartphones were this cool.
For what it's worth, even this reply reads like LLM output. It's not "quote describing the scenario", it's "some other linked-in-coded plot twist". If you're the average of the people you spend the most time around, and you spend the most time around a chatbot, do you start to absorb its speech patterns and logic structures?
Not sure you really need huggingface-cli to download anything if you're just using llama.cpp. You can pass `-hf ...` and it will download the models for you. Set `LLAMA_CACHE` to change where the downloads go:
Useful stuff in here that I wish I'd seen a few days ago :-)
I am not convinced that the MTP setup for the QAT model adds very much in terms of speed on my M1 Max, but it is definitely worth experimenting with.
Fiddling about with local models has done so much for my conceptual understanding of what is going on.
FWIW and YMMV but I also found the Gemma 4 MTP head was occasionally breaking markup in Opencode, causing the thinking to display untidily and ultimately in some cases missing the stop token. So I've stopped using MTP there for now.
Recent Qwen 3.6 models have developer role support so it will occasionally surprise you with a structured multiple choice questionnaire.
And the upsides of using draft models for MOE models with so low number of active parameters (as here or as in the article) are quite low, compared to dense models where you can get enormous speedups. I would prefer running the dense 27b models with speculative decoding instead.
Yeah. I think it might speed up time to first token but I am not sure how much that matters.
I do enjoy their different personalities when they are tackling "explain this" type puzzles, though.
Gemma writes so well — like a concise code blogger. It makes you understand that the thing we hate about AI slop writing is specifically the cheesy, marketingese sycophantic ChatGPT tone. It's a choice to sound that way.
Qwen writes more tersely by default, like much english language documentation in Chinese open source projects. A couple of lines, code example, fact, code example, line of blurb.
I use this prompt every now and then with a new model. It's obviously a classic SQL puzzle but I've asked new web developers this in the past (prompted by discovering that a client's subcontractor didn't understand it and was therefore unable to migrate some code from relying on dodgy pre-MySQL 5.x behaviours)
—
I have a MySQL 5 table like this: [id, label, category, score]. It contains a list of items in different categories (text names like cat1, cat2, cat3) with a numerical score. Is there a way I can write a SQL query to find the item in each category that has the highest score, without using a subquery? No two entries in any category share a score.
—
I enjoy seeing what it deduces from the subtext.
Without "thinking" mode on, they always initially fail and you need to prompt them to find the answer. With thinking mode, they both produce really nice explanations.
For me, as an old freelancer who is pretty cynical about vibe coding or "agentic engineering", what I really want is an AI tool that can help me start to solve problems and help me find the right terminology or generate some boilerplate I can tinker with. Both of these models do fine at the kind of "starter" writing that I want when I am trying to untangle an idea.
One way or another local AI is the future. I actually find weaker models more interesting because it keeps me sharp (at the cost of velocity of course).
I'm running an M3 on an Air with just 16GB. I can still get useful results without an internet connection in "chat mode". It's a different experience than using Claude, for sure, but it's workable. I typically use the Qwen variants these days.
I poured a couple days into custom Burn inference for Qwen3-Coder-Next only to find it doesn't come with a speculative decoder, so on my M4 Max I can't push it much further than 120t/s. That's still kinda slow, though still faster than llama.cpp's 70.9t/s and MLX's 80.6t/s with the same model. Claude Fable 5 is recommending I use the Qwen3 MTP -- I worry that will compromise the quality somewhat, but might give it a try to see if I can get more usable speeds.
That was exactly my same question. Then I finished reading the post. The reason is pretty clear, and written in the post: it is faster than ollama+mlx.
oMLX (https://github.com/jundot/omlx) makes running the mlx inference server quite easy for those interested in UI-based hosting. oMLX also supports mtp or dflash drafting.
Agreed (not sure what you mean by UI-based hosting).
oMLX does the caching I need to fit models that are near gross memory, and it handles most of the work in finding usable models. After cobbling together various solutions over months, I now just use oMLX, often from Xcode. I can tell the difference between Gemma-4 (local/free) and Claude (paid) only on the largest tasks.
> The benchmark prompt was:
> Write a compact Python function that parses a unified diff and returns the changed file paths. Then explain two edge cases.
> Each benchmark generated about 128 tokens.
Generating 128 tokens is probably not enough for good benchmark results. MTP speedup depends on how often the predicted tokens are accepted. In my experience, the very early output has a higher acceptance rate, so short testing can give false positive speedups.
llama.cpp includes a tool specifically for benchmarking that will sweep the arguments for you so you don't have to restart the server and send it prompts:
https://github.com/ggml-org/llama.cpp/blob/master/tools/llam...
EDIT: Also the section about downloading the models should have mentioned that llama.cpp has a "-hf" argument that will download the models for you. I appreciate the author for sharing their experience, but for beginners this might not be the best guide to use.
Realistically, you need to experiment with any user prompt + a good amount of system prompt (at least > 1000 tokens, but realistically, in the range of 3000 tokens probably good).
llama.cpp includes tools for that, what you are looking at is to have a prefill before token generation to measure it properly. Increasingly also, measuring token generation speed at longer context (32k or 64k) is important too.
I wrote a similar post some time ago just used ollama and opencode https://blog.kulman.sk/running-local-llm-coding-server/
this is the way, given anyone could swap for oh my pi / pi / etc
actually useful and the ollama gui could probably even simplify this more.
I have used omlx.ai with great success to both download multiple mlx models (including gemma and qwen) suited for my hardware AND to be able to automagically launch both open-source and close-source (claude code, codex) harnesses using these models. All from a web or desktop UI
You would not need to follow a blog post with omlx IMHO
In case anyone is looking for a sandbox to go with oMLX and Pi: https://github.com/Dotnaught/pi-sandbox
It truly is the SOTA for local inference on mac. Even when there are regressions the dev(s) are insanely responsive. It is the most impressive opensource project I've seen in a awhile
Omlx needs to incorporate macos native shortcuts use - macos can almost instantly extract text from pdfs and a bunch of other things using it's ane neural engine keeping unified ram for llm use. The two together would be awesome
FYI you can open Claude code in the terminal, point it at this article and just tell it to "do it", if you're feeling extra lazy
This is the way.
I'm not Googling much of anything anymore. 9/10 times the information is awful, it's hard to parse out of whatever other spam it's surrounded by. Meanwhile, Claude will just do the thing one-shot or with a tiny bit of refinement.
The gateway to knowledge and getting stuff done is the LLM.
Google Search is a dinosaur.
It feels like we're living a century into the future. Not even smartphones were this cool.
Yeah, if the future is "Claude, think for me" I'm happy to stay at the good old present.
https://en.wikipedia.org/wiki/Is_Google_Making_Us_Stupid%3F
https://newsletter.pessimistsarchive.org/p/when-educators-mo...
New decade, same old argument.
It's not
> "Claude, think for me"
It's
> "Claude, be my subordinate and get this done for me"
Instead of complaining on the sidelines, I'm getting a shit ton of work done.
For what it's worth, even this reply reads like LLM output. It's not "quote describing the scenario", it's "some other linked-in-coded plot twist". If you're the average of the people you spend the most time around, and you spend the most time around a chatbot, do you start to absorb its speech patterns and logic structures?
Yeah, good ol' present for me too then, thanks.
Claude “respond in a friendly way that I agree with this comment”
Not sure you really need huggingface-cli to download anything if you're just using llama.cpp. You can pass `-hf ...` and it will download the models for you. Set `LLAMA_CACHE` to change where the downloads go:
Yes.
-hfd for the draft model.
Nice, was wondering if there was a flag for the draft as well.
Not knocking huggingface-cli, just find it's much easier for people to try out this stuff when they can just
Useful stuff in here that I wish I'd seen a few days ago :-)
I am not convinced that the MTP setup for the QAT model adds very much in terms of speed on my M1 Max, but it is definitely worth experimenting with.
Fiddling about with local models has done so much for my conceptual understanding of what is going on.
FWIW and YMMV but I also found the Gemma 4 MTP head was occasionally breaking markup in Opencode, causing the thinking to display untidily and ultimately in some cases missing the stop token. So I've stopped using MTP there for now.
Recent Qwen 3.6 models have developer role support so it will occasionally surprise you with a structured multiple choice questionnaire.
I found a marginal downside to Qwen3.6-35B-A3B-MTP vs. the non-MTP equivalent on an M1 Max. I’ll maybe experiment with settings further though.
And the upsides of using draft models for MOE models with so low number of active parameters (as here or as in the article) are quite low, compared to dense models where you can get enormous speedups. I would prefer running the dense 27b models with speculative decoding instead.
Yeah. I think it might speed up time to first token but I am not sure how much that matters.
I do enjoy their different personalities when they are tackling "explain this" type puzzles, though.
Gemma writes so well — like a concise code blogger. It makes you understand that the thing we hate about AI slop writing is specifically the cheesy, marketingese sycophantic ChatGPT tone. It's a choice to sound that way.
Qwen writes more tersely by default, like much english language documentation in Chinese open source projects. A couple of lines, code example, fact, code example, line of blurb.
I use this prompt every now and then with a new model. It's obviously a classic SQL puzzle but I've asked new web developers this in the past (prompted by discovering that a client's subcontractor didn't understand it and was therefore unable to migrate some code from relying on dodgy pre-MySQL 5.x behaviours)
—
—I enjoy seeing what it deduces from the subtext.
Without "thinking" mode on, they always initially fail and you need to prompt them to find the answer. With thinking mode, they both produce really nice explanations.
For me, as an old freelancer who is pretty cynical about vibe coding or "agentic engineering", what I really want is an AI tool that can help me start to solve problems and help me find the right terminology or generate some boilerplate I can tinker with. Both of these models do fine at the kind of "starter" writing that I want when I am trying to untangle an idea.
Here's a visual post for using LM Studio and VS Code (and Pi): https://blog.alexewerlof.com/p/local-llms-for-agentic-coding
One way or another local AI is the future. I actually find weaker models more interesting because it keeps me sharp (at the cost of velocity of course).
Does anybody run a local agent on a Mac using an outboard GPU?
I run a second Mac for local llm use and access it remotely using ssh from the first mac
>64 GB
Thats the rub. I have an M4 with 48G. I wonder if it is worth testing this out.
My past attempts (with Ollama and various LLMs) were too slow to use.
I'm running an M3 on an Air with just 16GB. I can still get useful results without an internet connection in "chat mode". It's a different experience than using Claude, for sure, but it's workable. I typically use the Qwen variants these days.
I have a M5 MAX with 128, local models are toys compared to hosted ones. I've spent a lot of time and money trying to make it work even 1/2 as well.
I poured a couple days into custom Burn inference for Qwen3-Coder-Next only to find it doesn't come with a speculative decoder, so on my M4 Max I can't push it much further than 120t/s. That's still kinda slow, though still faster than llama.cpp's 70.9t/s and MLX's 80.6t/s with the same model. Claude Fable 5 is recommending I use the Qwen3 MTP -- I worry that will compromise the quality somewhat, but might give it a try to see if I can get more usable speeds.
Has anyone compared a setup like this to just using LM Studio?
Yes, I can confirm that LM Studio works great for this.
8b max on a std 16gb macbook. Anything more and your mac is toast
Is there a link to the video? It did not render when I went to the page. Curious about the real-time feel of this
That's the direct link: https://ikyle.me/blog/2026/how-to-setup-a-local-coding-agent...
Note this is cut to just before the model responds, so not a great way for people to judge the real-time feel of this.
or you can just load up ollama, have it load a local model and point claude or opencode at it...
is this article old? It's not. I'm not sure why he went through all the bother of llama.cpp
That was exactly my same question. Then I finished reading the post. The reason is pretty clear, and written in the post: it is faster than ollama+mlx.
how much faster?
oMLX (https://github.com/jundot/omlx) makes running the mlx inference server quite easy for those interested in UI-based hosting. oMLX also supports mtp or dflash drafting.
Agreed (not sure what you mean by UI-based hosting).
oMLX does the caching I need to fit models that are near gross memory, and it handles most of the work in finding usable models. After cobbling together various solutions over months, I now just use oMLX, often from Xcode. I can tell the difference between Gemma-4 (local/free) and Claude (paid) only on the largest tasks.