Yes, just — 9.8 GB spare
Qwen3 235B-A22B at Q4_K_M fits your M2 Ultra · 192 GB entirely in unified memory at 8K context, at an estimated 17 tokens per second. Past 61K the KV cache pushes it over — quantise the cache to q8_0, or step down a quantisation, to go longer.
Fully in unified memory
8K context
Q4_K_M · 132.1 GB
Apache 2.0
Released Apr 2025
Workstation class. Realistically a 192 GB unified-memory or multi-GPU model.
The VRAM budget
weights 132.1 GB
Weights 132.1 GB
KV cache @ 8K 1.47 GB
Runtime overhead 0.6 GB
Free 9.8 GB of 144.0 GB
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| Q8_0 | 232.5 GB | 234.6 GB | — | ~9.6 | −0.1% ppl | 90.6 GB over |
| Q6_K | 179.5 GB | 181.5 GB | — | ~12 | −0.4% ppl | 37.5 GB over |
| Q5_K_M | 155.1 GB | 157.2 GB | — | ~14 | −0.8% ppl | 13.2 GB over |
| Q4_K_M | 132.1 GB | 134.2 GB | 61K | 17 | −1.9% ppl | Recommended |
| Q3_K_M | 107.0 GB | 109.0 GB | 128K | 21 | −5.4% ppl | Long context |
| Q2_K | 91.6 GB | 93.7 GB | 128K | 24 | −15% ppl | Long context |
Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it.
How to run it
$ pip install mlx-lm $ mlx_lm.generate --model mlx-community/Qwen3-235B-A22B-4bit \ --max-tokens 512 --prompt "Hello"
Apple's own array framework. The fastest path on Apple Silicon. More on MLX.
01Download is 132.1 GB. Keep it on an SSD — a first load off a spinning disk takes minutes.
02macOS caps what the GPU may wire down at about 144.0 GB of 192 GB. Raising it with iogpu.wired_limit_mb is possible, and risky.
03Only 9.8 GB is spare, so a long system prompt can still push it over. Its real ceiling here is 61K context.