Can the RTX 4070 run InternLumina-U2?
InternLumina-U2·16B·GeForce RTX 4070·12 GB VRAM
Yes, but only at Q3_K_M (about 9.6 GB of 12 GB), which costs noticeable quality, at roughly 35 tokens/s.
- best quant
- Q3_K_M
- needs
- 9.6 GB
- usable
- 12 GB
- est. speed
- ~35 tok/s
Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the InternLumina-U2 page →
Every quantisation of InternLumina-U2 on the RTX 4070
| Quant | Memory | Fit | Speed |
|---|---|---|---|
| Q2_K | 7.1 GB | fits | ~53 tok/s |
| Q3_K_M pick | 9.6 GB | fits | ~35 tok/s |
| Q4_K_M | 11 GB | tight | ~29 tok/s |
| Q5_K_M | 13 GB | offload | ~15 tok/s |
| Q6_K | 15 GB | offload | ~8.7 tok/s |
| Q8_0 | 19 GB | offload | ~4.5 tok/s |
| FP16 / BF16 | 33 GB | offload | ~1.5 tok/s |
Speed is a bandwidth-based estimate for generating text (504 GB/s); prompt processing and long contexts are slower.
Measured speed on the RTX 4070
No community reports for this pair yet.
Report your own speed →Omni models that fit the RTX 4070 comfortably
All that fit →| Model | Params | Best quant | Memory | Speed |
|---|---|---|---|---|
| 12B | Q4_K_M | ≈10 GBFits in memory | ~38 tok/s | |
| 14.7B | Q5_K_M | ≈11 GBFits in memory | ~27 tok/s | |
| 10.7B | Q8_0 | ≈8.9 GBFits in memory | ~24 tok/s | |
| 12B | Q4_K_M | ≈11 GBFits in memory | ~38 tok/s | |
| 9.4B | Q8_0 | ≈10 GBFits in memory | ~28 tok/s | |
| 16.3BA1.5B | Q4_K_M | ≈10 GBFits in memory | ~305 tok/s |
Other GPUs that run InternLumina-U2
Frequently asked questions
- Can the RTX 4070 run InternLumina-U2?
- Yes, but only at Q3_K_M (about 9.6 GB of 12 GB), which costs noticeable quality, at roughly 35 tokens/s.
- How fast does InternLumina-U2 run on the RTX 4070?
- An estimate from memory bandwidth gives about 35 tokens/s at Q3_K_M.
- Which quantisation of InternLumina-U2 should I use on the RTX 4070?
- Q3_K_M: the largest one that fits. Expected quality: noticeable loss 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.