Can the Instinct MI50 32GB run WeMM-Video-35BA3B?
WeMM-Video-35BA3B·36.2B (5B active)·Instinct MI50 32GB·32 GB VRAM
Yes. The Instinct MI50 32GB runs WeMM-Video-35BA3B fully in VRAM at Q6_K (about 30 GB of 32 GB), at roughly 138 tokens/s.
- best quant
- Q6_K
- needs
- 30 GB
- usable
- 32 GB
- est. speed
- ~138 tok/s
Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the WeMM-Video-35BA3B page →
Every quantisation of WeMM-Video-35BA3B on the Instinct MI50 32GB
| Quant | Memory | Fit | Speed |
|---|---|---|---|
| Q2_K | 13 GB | fits | ~345 tok/s |
| Q3_K_M | 18 GB | fits | ~232 tok/s |
| Q4_K_M | 23 GB | fits | ~187 tok/s |
| Q5_K_M | 26 GB | fits | ~159 tok/s |
| Q6_K pick | 30 GB | fits | ~138 tok/s |
| Q8_0 | 39 GB | offload | ~28 tok/s |
| FP16 / BF16 | 72 GB | too big | — |
Speed is a bandwidth-based estimate for generating text (1024 GB/s); prompt processing and long contexts are slower.
Measured speed on the Instinct MI50 32GB
No community reports for this pair yet.
Report your own speed →Vision models that fit the Instinct MI50 32GB comfortably
All that fit →| Model | Params | Best quant | Memory | Speed |
|---|---|---|---|---|
| 36BA3B | Q6_K | ≈30 GBFits in memory | ~228 tok/s | |
| 36BA3B | Q6_K | ≈29 GBFits in memory | ~228 tok/s | |
| 31.3B | Q5_K_M | ≈29 GBFits in memory | ~25 tok/s | |
| 27.8B | Q6_K | ≈25 GBFits in memory | ~25 tok/s | |
| 35.2BA3B | Q6_K | ≈29 GBFits in memory | ~228 tok/s | |
| 29.8B | Q8_0 | ≈30 GBFits in memory | ~18 tok/s |
Other GPUs that run WeMM-Video-35BA3B
Frequently asked questions
- Can the Instinct MI50 32GB run WeMM-Video-35BA3B?
- Yes. The Instinct MI50 32GB runs WeMM-Video-35BA3B fully in VRAM at Q6_K (about 30 GB of 32 GB), at roughly 138 tokens/s.
- How fast does WeMM-Video-35BA3B run on the Instinct MI50 32GB?
- An estimate from memory bandwidth gives about 138 tokens/s at Q6_K.
- Which quantisation of WeMM-Video-35BA3B should I use on the Instinct MI50 32GB?
- Q6_K: the largest one that fits. Expected quality: near-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.