What people run, by hardware
Pick the kind of machine you have. You'll see the model sizes people report running on it, most reported first, with the speed they typically get. It tells you what works on hardware like yours. It doesn't tell you which model gives the best answers.
To check a specific model on your exact card, use the calculator. For coding, see the coding models by VRAM.
A Mac counts when its report names the product, or when only that product was sold with its chip and memory.
Apple Silicon 48-96 GB· MacBook Pro only
70B-class models fit at 4 bits. Memory speed 273 to 614 GB/s across the 10 devices with reports, so a typical figure here spans all of them. 26 single-device reports.
| # | Model | Reports | Gen t/s | PP t/s | With speculation |
|---|---|---|---|---|---|
| 1 | Qwen3.827B Alibaba · mostly 4-bit No generation data · with speculation 60 t/s (7) | 8 | no generation data | no PP data | 60 7 runs |
| 2 | Qwen3.8 Flash-Next125B · 6B active Alibaba · mostly Q4_0 No generation data | 4 | no generation data | no PP data | none |
| 3 | Qwen3.627B Alibaba · mostly 4-bit Gen 20 t/s (2 runs) · PP 121 t/s (1 run) · with speculation 63 t/s (1) | 3 | 20 2 runs, 2 devices | 121 1 run | 63 1 run |
| 4 | DeepSeek V4 Flash284B · 13B active DeepSeek · mostly Q2-Q4 mixed imatrix Gen 13 t/s (1 run) | 2 | 13 1 run | no PP data | none |
| 5 | Kimi K32800B · 104B active Moonshot AI · mostly VQ3R No generation data | 2 | no generation data | no PP data | none |
| 6 | Qwen3.635B · 3B active Alibaba · mostly 4-bit group-64 No generation data · with speculation 108 t/s (2) | 2 | no generation data | no PP data | 108 2 runs |
| 7 | Qwen3.8 Uncensored27B Alibaba · mostly 4-bit No generation data · with speculation 59 t/s (2) | 2 | no generation data | no PP data | 59 2 runs |
| 8 | Gemma 426B · 4B active Google DeepMind Gen 97 t/s (1 run) | 1 | 97 1 run | no PP data | none |
| 9 | DeepSeek V4 Flash 9B distill9B DeepSeek · mostly Q4_K_M Gen 44 t/s (1 run) | 1 | 44 1 run | no PP data | none |
| 10 | Qwen3.527B Alibaba · mostly affine 4bit No generation data · PP 461 t/s (1 run) | 1 | no generation data | 461 1 run | none |
How this list works
Each row is one model size, because a 27B and a 397B from the same family are different choices. Rows are ordered by how many people report running that size on this hardware.
Generation speed is the median of runs on one device, serving one request, without speculative decoding. PP is the median prompt-processing speed from plain runs that reported it, so its run count can be smaller. Speculative decoding (MTP, a draft model, DFlash, n-gram lookup) can make the same model several times faster, so those runs get their own column. A report that doesn't make clear which kind of run it was counts in neither.
Hardware is grouped by how much it holds and how fast its memory is, since memory speed sets how fast a model writes. Click a model for every report behind its row.