Can Falcon 40B run on the Mac M2 Ultra 192GB?
Yes. At 8K context the best fit is F16, using about 79.4 GB of the 144 GB usable on the Mac M2 Ultra 192GB. Estimated speed ceiling: ~10 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | 15.7 GB Fits | 16.1 GB Fits | 19.0 GB Fits | 30.2 GB Fits | 51 |
| 3-bit (Q3_K_M) • | 19.7 GB Fits | 20.2 GB Fits | 23.0 GB Fits | 34.2 GB Fits | 40 |
| 4-bit (Q4_K_M) • | 24.7 GB Fits | 25.2 GB Fits | 28.0 GB Fits | 39.3 GB Fits | 31 |
| 5-bit (Q5_K_M) • | 29.5 GB Fits | 30.0 GB Fits | 32.8 GB Fits | 44.1 GB Fits | 26 |
| 6-bit (Q6_K) • | 33.1 GB Fits | 33.6 GB Fits | 36.4 GB Fits | 47.6 GB Fits | 23 |
| 8-bit (Q8_0) | 42.4 GB Fits | 42.9 GB Fits | 45.7 GB Fits | 56.9 GB Fits | 18 |
| 16-bit (F16/BF16) | 78.9 GB Fits | 79.4 GB Fits | 82.2 GB Fits | 93.4 GB Fits | 10 |
• real GGUF file size from mradermacher/falcon-40b-i1-GGUF. Speed is a bandwidth ceiling (800 GB/s), not a benchmark. Methodology.
What fits instead
- Cheapest listed GPU that runs Falcon 40B: RTX 5090 (32 GB, launch MSRP $1,999) at Q4_K_M, fully in VRAM at 8K context.
- Largest popular model that fits the Mac M2 Ultra 192GB: DeepSeek V4 Flash 0731 (304B, MoE) at F16, about 11.9 GB at 8K context.
"Runs" and "fits" mean the whole model in GPU memory at 8K context (4-bit or better for the model pick). Prices are US launch MSRPs, not today's street prices. "Popular" means the 60 most-downloaded models we track.
More
Falcon 40B on every GPU and Mac · Every model that runs on the Mac M2 Ultra 192GB · Apple Silicon guide · Try other settings in the calculator
Guides: Best GPU for local LLMs by budget · Can I run gpt-oss, Llama 70B or Qwen3? · Mac vs PC for local LLMs
Data: Hugging Face Hub (tiiuae/falcon-40b), Apple specs. Model data last checked 2026-11-17; refreshed hourly.