Can I Run Cosmos 3 Super on Apple M2 Ultra?
Written by Jakub Rusinowski · Last updated July 21, 2026
Yes, but it's tight. Cosmos 3 Super needs 131.1 GB at 8K context (128 GB weights + 3.1 GB KV cache/overhead) at BF16 and the Apple M2 Ultra has 144 GB usable, leaving about 12.9 GB of headroom, at roughly 7 tok/s (estimated).
See what else this hardware can run →
The numbers
| Model | Cosmos 3 Super |
| Parameters | 64B |
| Quantization | BF16 |
| VRAM needed (8K context) | 131.1 GB |
| Apple M2 Ultra usable memory | 144 GB |
| Verdict | Runs |
| Estimated speed | ~7 tok/s |
Cosmos 3 Super on the Apple M2 Ultra: VRAM by quantization
| Quant | VRAM needed | Fits 192 GB? | Max context | Whole PC |
|---|---|---|---|---|
| F16 | 131.1 GB | ✓ Yes | 16K | — |
| Q8_0 | 71.1 GB | ✓ Yes | 16K | — |
| Q6_K | 55.6 GB | ✓ Yes | 16K | — |
| Q5_K_M | 48.5 GB | ✓ Yes | 16K | — |
| Q4_K_M | 41.8 GB | ✓ Yes | 16K | — |
| Q3_K_M | 30.4 GB | ✓ Yes | 16K | — |
| Q2_K | 24.2 GB | ✓ Yes | 16K | — |
Assumes an 8K-token context with an f16 KV cache, on an estimated architecture — this model publishes no config we can read. A longer window needs more; a quantized KV cache needs less. “Max context” is the largest window that still fits in 192 GB of unified memory. Figures are estimates from parameter count, quantization and memory bandwidth — the analyzer lets you tune KV-cache quant and context.
Context costs VRAM too. At BF16 on the Apple M2 Ultra: 8K 131.1 GB · 16K 133.4 GB. Every window this model publishes still fits 192 GB.
What to do next
Nothing to change — run it at BF16. If you want more capability, check whether the next size up also fits.
No compatible GPU? Cosmos 3 Super on 64 GB of system RAM, CPU only: runs comfortably.
Running Cosmos 3 Super on this card:
Cosmos 3 Super needs ~131 GB but this GPU has 192 GB. Rent a 2× A100 (160 GB)-class GPU by the hour instead of buying one:
Affiliate links — we may earn a commission if you sign up, at no extra cost to you.
Cloud rates verified 2026-07 — estimates, and marketplace prices vary. Buying price is GPU MSRP only, not a full PC.
Other Cosmos 3 sizes on the Apple M2 Ultra
- Cosmos 3 Nano — runs (34.2 GB)
- Cosmos 3 Edge — runs (9.7 GB)
Cosmos 3 Super on nearby hardware
- Apple M3 Max (128 GB) — runs at a lower quant
All Cosmos 3 sizes on the Apple M2 Ultra | Cosmos 3 Super model page | Apple M2 Ultra GPU page | VRAM calculator | Check your hardware