Can the Apple M4 (16 GB) run Crowd-v1?
Crowd-v1·1B·Apple M4 (16 GB)·16 GB unified memory
Yes. The Apple M4 (16 GB) runs Crowd-v1 fully in unified memory at FP16 / BF16 (about 2.8 GB of 11 GB), at roughly 33 tokens/s.
- best quant
- FP16 / BF16
- needs
- 2.8 GB
- usable
- 11 GB
- est. speed
- ~33 tok/s
Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the Crowd-v1 page →
Every quantisation of Crowd-v1 on the Apple M4 (16 GB)
| Quant | Memory | Fit | Speed |
|---|---|---|---|
| Q2_K | 1.2 GB | fits | ~201 tok/s |
| Q3_K_M | 1.3 GB | fits | ~135 tok/s |
| Q4_K_M | 1.4 GB | fits | ~109 tok/s |
| Q5_K_M | 1.6 GB | fits | ~93 tok/s |
| Q6_K | 1.7 GB | fits | ~80 tok/s |
| Q8_0 | 1.9 GB | fits | ~62 tok/s |
| FP16 / BF16 pick | 2.8 GB | fits | ~33 tok/s |
Speed is a bandwidth-based estimate for generating text (120 GB/s); prompt processing and long contexts are slower.
Measured speed on the Apple M4 (16 GB)
No community reports for this pair yet.
Report your own speed →Chat models that fit the Apple M4 (16 GB) comfortably
All that fit →| Model | Params | Best quant | Memory | Speed |
|---|---|---|---|---|
| 1B | Q8_0 | ≈1.8 GBFits in memory | ~62 tok/s | |
| 752M | Q8_0 | ≈2.0 GBFits in memory | ~83 tok/s | |
| 1.2B | Q8_0 | ≈2.0 GBFits in memory | ~50 tok/s | |
| 1.1B | Q8_0 | ≈1.7 GBFits in memory | ~56 tok/s | |
| 1.7B | Q8_0 | ≈2.7 GBFits in memory | ~36 tok/s | |
| 1.5B | Q8_0 | ≈2.6 GBFits in memory | ~40 tok/s |
Other GPUs that run Crowd-v1
Frequently asked questions
- Can the Apple M4 (16 GB) run Crowd-v1?
- Yes. The Apple M4 (16 GB) runs Crowd-v1 fully in unified memory at FP16 / BF16 (about 2.8 GB of 11 GB), at roughly 33 tokens/s.
- How fast does Crowd-v1 run on the Apple M4 (16 GB)?
- An estimate from memory bandwidth gives about 33 tokens/s at FP16 / BF16.
- Which quantisation of Crowd-v1 should I use on the Apple M4 (16 GB)?
- FP16 / BF16: the largest one that fits. Expected quality: lossless compared with the full-precision weights.
Memory and speed are estimates from the model's size and the card's memory bandwidth; real numbers depend on the runtime, drivers and context length. Community figures are medians of user reports.