Can I run Qwen3.6 35B-A3B on the Apple M4?
Runs slowlyQ4_K_M runs entirely from 22.5 GB of system RAM: Runs slowly, est. 25.4 tok/s (17.8–33.1, theoretical estimate ±30%).
32 GB unified memory · 120.0 GB/s · FP16 3.8 TFLOPS. Assumes 8k context and an f16 KV cache. Verdicts are shown for each memory size.
Verdict by memory size
The page uses the 32 GB configuration; here is every size.
- 16 GBWon't runQ4_K_M doesn't fit
- 24 GBWon't runTry IQ4_XS: Runs slowlyest. 28.5 tok/s (19.9–37.0, theoretical estimate ±30%)Q2_K (heavy quality loss): Runs great
- 32 GBRuns slowlyest. 25.4 tok/s (17.8–33.1, theoretical estimate ±30%)Try IQ4_XS: Runs great(used on this page)
Every quant of Qwen3.6 35B-A3B on the Apple M4
| Quant | File size | Verdict | Speed | Memory | Runs as | Notes |
|---|---|---|---|---|---|---|
| Q2_K | 13.51 GB | Runs great | est. 43.0 tok/s34.4–51.6calibrated estimate ±20% | 14.3 / 22.4 GB | Unified memory | — |
| IQ4_XS | 19.70 GB | Runs great | est. 38.0 tok/s30.4–45.6calibrated estimate ±20% | 20.4 / 22.4 GB | Unified memory | — |
| Q4_K_Mbaseline | 22.29 GB | Runs slowly | est. 25.4 tok/s17.8–33.1theoretical estimate ±30% | 22.5 GB RAM | CPU only | CPU inference — capped at Runs slowly |
| Q8_0 | 37.81 GB | Won't run | — | Needs about 38.6 GB; 28.0 GB of memory is free (22.4 GB usable by the GPU) | — | — |
Models with a thinking mode spend extra tokens thinking, so their speed thresholds are 1.5× stricter (30 / 12 / 3 tok/s). With thinking turned off, a model may do one grade better.
Where the memory goes at Q4_K_M
- Weights in system RAM
- 22.5 GB
Context length vs. KV cache
| Context | KV f16 | KV q8_0 | KV q4_0 |
|---|---|---|---|
| 4k | 22.4 GBRuns slowly est. 26.5 tok/s18.6–34.5theoretical estimate ±30% | 22.3 GBRuns slowly est. 27.1 tok/s19.0–35.2theoretical estimate ±30% | 22.3 GBRuns slowly est. 27.4 tok/s19.2–35.6theoretical estimate ±30% |
| 8k | 22.5 GBRuns slowly est. 25.4 tok/s17.8–33.1theoretical estimate ±30% | 22.4 GBRuns slowly est. 26.5 tok/s18.5–34.4theoretical estimate ±30% | 22.3 GBRuns slowly est. 27.0 tok/s18.9–35.2theoretical estimate ±30% |
| 16k | 22.6 GBRuns slowly est. 23.5 tok/s16.5–30.6theoretical estimate ±30% | 22.5 GBRuns slowly est. 25.3 tok/s17.7–32.9theoretical estimate ±30% | 22.4 GBRuns slowly est. 26.4 tok/s18.5–34.3theoretical estimate ±30% |
| 32k | 23.0 GBRuns slowly est. 20.4 tok/s14.3–26.5theoretical estimate ±30% | 22.6 GBRuns slowly est. 23.3 tok/s16.3–30.2theoretical estimate ±30% | 22.5 GBRuns slowly est. 25.2 tok/s17.6–32.7theoretical estimate ±30% |
| 64k | 23.6 GBRuns slowly est. 16.1 tok/s11.3–20.9theoretical estimate ±30% | 23.0 GBRuns slowly est. 20.0 tok/s14.0–26.0theoretical estimate ±30% | 22.7 GBRuns slowly est. 23.0 tok/s16.1–29.9theoretical estimate ±30% |
| 128k | 25.0 GBRuns slowly est. 11.4 tok/s7.9–14.8theoretical estimate ±30% | 23.7 GBRuns slowly est. 15.7 tok/s11.0–20.4theoretical estimate ±30% | 23.1 GBRuns slowly est. 19.7 tok/s13.8–25.6theoretical estimate ±30% |
| 256k | 27.7 GBRuns slowly est. 7.1 tok/s5.0–9.3theoretical estimate ±30% | 25.2 GBRuns slowly est. 10.9 tok/s7.6–14.2theoretical estimate ±30% | 23.8 GBRuns slowly est. 15.2 tok/s10.7–19.8theoretical estimate ±30% |
S = Runs great · A = Runs well · B = Runs slowly · F = Won't run
At Q4_K_M the verdict stays Runs slowly up to 256k context with an f16 KV cache, and up to 256k with q8_0.
Estimated speed
- Generation
- est. 25.4 tok/s (17.8–33.1, theoretical estimate ±30%)
- Prompt processing
- Prompt-processing estimates only apply when the whole model runs on the GPU.
Why this verdict: CPU inference — capped at Runs slowly
Measured on this exact combination
No public measurement for this exact combination yet; the numbers above are estimates.
If this is not enough
A smaller model that runs well on the Apple M4
Runs greatMiniCPM-V 4.7 35B-A3B: Runs great, est. 37.5 tok/s (30.0–45.0, calibrated estimate ±20%)
A lower quant of Qwen3.6 35B-A3B
Runs greatIQ4_XS (19.70 GB): Runs great, est. 38.0 tok/s (30.4–45.6, calibrated estimate ±20%)
Qwen3.6 35B-A3B on the Apple M4 Pro
Runs greatQ4_K_M: Runs great, est. 68.8 tok/s (55.0–82.5, calibrated estimate ±20%)
How to run it
Commands for the Q4_K_M file. Both tools download from Hugging Face on first run.
File: Qwen_Qwen3.6-35B-A3B-Q4_K_M.gguf (22.29 GB) from bartowski/Qwen_Qwen3.6-35B-A3B-GGUF.
llama-server -hf bartowski/Qwen_Qwen3.6-35B-A3B-GGUF:Q4_K_M -c 8192 -ngl 0 -fa on- -ngl 0 keeps everything on the CPU.
- -fa on enables flash attention (needed for KV cache quantization).
- --cache-type-k q8_0 --cache-type-v q8_0 shrinks the KV cache to 0.09 GB at 8k context.
Terminal:
OLLAMA_CONTEXT_LENGTH=8192 ollama serve
ollama run hf.co/bartowski/Qwen_Qwen3.6-35B-A3B-GGUF:Q4_K_M- Ollama defaults to 4k context on GPUs under 24 GB. Set OLLAMA_CONTEXT_LENGTH when starting the server, or /set parameter num_ctx inside the chat.
/set parameter num_ctx 8192
Related pages
Guide: Running local LLMs on a Mac (M4, M5): how much unified memory?
Other models on the Apple M4
Qwen3.6 35B-A3B on other hardware
Frequently asked questions
Can I run Qwen3.6 35B-A3B on the Apple M4?
Q4_K_M runs entirely from 22.5 GB of system RAM: Runs slowly, est. 25.4 tok/s (17.8–33.1, theoretical estimate ±30%). Q4_K_M needs 22.5 GB at 8k context; the GPU can use 22.4 GB of the unified memory.
How much context can Qwen3.6 35B-A3B use on the Apple M4?
At Q4_K_M the verdict stays Runs slowly up to 256k context with an f16 KV cache (5.4 GB of KV) and up to 256k with a q8_0 KV cache (2.9 GB). The model supports up to 256k.
Which quant should I use, and how fast is it?
Q4_K_M (22.29 GB) is the recommended balance: Runs slowly, est. 25.4 tok/s (17.8–33.1, theoretical estimate ±30%). It is also the largest tracked file that gets this verdict.
Speeds are estimates from memory bandwidth, calibrated against public benchmarks, and each one comes with an error band and a confidence label. Real results vary with drivers, backend, context length and thermals.