Model reference · open weights
xflux_text_encoders is an open-weight language model from XLabs-AI. xflux_text_encoders (BF16) weighs 9.5 GB; the smallest configuration that runs it is RTX 4060 Ti 16 GB.
What it is
| Released by | XLabs-AI |
|---|---|
| Released | 2024-08-11 |
| Parameters | 4.8B |
| VRAM | 9.5 GB for the weights |
What it runs on
| Card | Requests at once | Context max | Memory | |
|---|---|---|---|---|
| 8K each | 32K each | |||
| RTX 3060 12 GB | — | — | 3K | 11.6 GB |
| RTX 4060 Ti 16 GB | 1 | — | 12K | 15.4 GB |
| RTX 3090 24 GB | 4 | 1 | 32K | 23.4 GB |
| RTX 4090 24 GB | 4 | 1 | 32K | 23.4 GB |
| RTX 5090 32 GB | 6 | 1 | 51K | 31.0 GB |
| L40S 48 GB | 10 | 2 | 83K | 44.0 GB |
| A100 80 GB | 21 | 5 | 128K | 78.2 GB |
| H100 80 GB | 20 | 5 | 128K | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | 25 | 6 | 128K | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | 29 | 7 | 128K | 107 GB |
| H200 141 GB | 38 | 9 | 128K | 138 GB |
| B200 180 GB | 50 | 12 | 128K | 176 GB |
| Requests at once | 8K tokens each | 32K tokens each |
|---|---|---|
| 1 | 13.4 GB | 23.1 GB |
| 5 | 26.3 GB | 74.6 GB |
| 8 | 36.0 GB | 113 GB |
| 16 | 61.7 GB | 216 GB |
| 32 | 113 GB | 423 GB |
| 64 | 216 GB | 835 GB |
One card, with vLLM's small-card settings.
From the model card
Text encoder weights from Google's T5 model
Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.