6,920 karma · joined January 1, 2010
Non-NVIDIA backends tend to get less support and new features land slower, or features that are expected to improve performance wind up hurting it instead. That sort of thing.
For basic “token in/token out” workloads without fine tuning, it’s probably fine ??
Wouldn’t you be more surprised / concerned if the model refused to to criticize Google?
I've certainly used these models without wifi without any differences.
- A 27B "dense" model
- A 35B "Mixture of Experts" model, which activates only 3B parameters for each token.
For your hardware, I strongly recommend `unsloth/Qwen3.6-35B-A3B-GGUF:Q4_K_M`. I have an M1 Max with 32GB VRAM from 2021 that can read at ~300-500 tokens/sec and write at ~30 tokens/sec with llama-cpp's default settings, which is plenty fast. The 27B model can read ~70tok/sec and write ~5tok/sec.
The 35B MoE model technically takes slightly more memory but is much faster because it's doing 1/9th the work. It's not quite as "smart", but it's comparable.
/Users/gcr/llama.cpp/build/bin/llama-server
-hf unsloth/Qwen3.6-35B-A3B-GGUF:Q4_K_M
--no-mmproj-offload
--fit on
-c 65536 # edit to taste
--reasoning on --chat-template-kwargs '{"preserve_thinking": true}'
--sleep-idle-seconds 90 # very aggressive: purge model from vram after this long
-ctk q8_0 -ctv q8_0 # Optional. Lower memory use, but lower speed. Omit if you can.
I don't recommend ollama or lm-studio. Ollama's in the process of switching from their llama-cpp backend anyway, but their new go framework frequently OOMs and crashes on my hardware. I also don't recommend MLX-based inference backends on this hardware; I've found them to consistently reduce performance, contrary to what I've read online. I've tried all the llama-cpp metal forks, but right now, MTP, TurboQuant, MLX, etc etc etc are too new and just slow things down. It's all dust in the wind still.For agent harnesses, opencode is okay, as is pi or even Zed's built in agent panel. Claude code "works" with ANTHROPIC_BASE_URL=http://localhost:8080/v1, but is very chatty (the default system prompt burns 20k tokens). Crush (from the charm-bracelet folks) is particularly nice when starting out. I've personally converged on pi-agent under an otherwise-mostly-default setup. You can ask qwen to customize pi or write you an extension which helps a little.
You'll need to add `http://localhost:8080/v1` as an OpenAI-compatible model provider in your coding harness with any API key (doesn't matter) and any model identifier (doesn't matter with llama-cpp).
Note that pi doesn't have permissions. Everything is permitted. The hundred hungry ghosts you've trapped in a jar WILL find a way to delete your home folder someday. That's what Man gets for summoning demons without casting a circle of protection first. Flying too close to the sun etc etc etc
Take backups and then go have fun. Hope this helps.
I’m on an M1 Max with 32GB VRAM, so I’m looking forward to the 27B or 35B-A3B models. Is dropping $5k for an RTX 6000 or a DGX Spark really the best option?
This is an excellent closing statement.
I have only my own personal experience for frontier models, but I have seen different performance of Opus when used from Pi or Claude Code or Zed for example.
For others who are lacking context :-)
Getting any smart model to take a look at the task is the sort of lift that the speaker is usually pointing to.
They even send the “you’re being fired” email to their personal email they have on file. Didn’t even schedule a meeting.
Subscription usage refills every five hours. API billing doesn't.
Hoping the Zed team will clarify.
I have several libraries that I've written. Not one single serious security bug in them has been found since 1991. Granted, nobody uses my libraries...
Not to diminish your team's achievement! :D But it's important to contextualize claims like this with information about what your userbase looks like