Best GPU for FLUX.1 [schnell]: VRAM Requirements
FLUX.1 [schnell] needs about 37.2 GB of GPU memory in BF16 and 9.3 GB at 4 bits. See which GPUs fit it for inference and fine-tuning.
FLUX.1 [schnell] needs about 37.2 GB of VRAM in FP16, 18.6 GB in FP8 and 9.3 GB at 4 bits for inference. In FP16 it fits on one NVIDIA L40S 48GB; a 4-bit build fits on one NVIDIA GeForce RTX 5080 16GB.
FLUX.1 [schnell] is a 16.9B-parameter image generation model from Black Forest Labs, released in August 2024 and published under the Apache 2.0. Its weights alone take 33.8 GB in BF16 and 8.5 GB at 4 bits per parameter; the tables below add runtime memory and fine-tuning, then list the GPUs that hold each setup.
Recommended GPUs for FLUX.1 [schnell]
Best for inference
1 GPU × NVIDIA L4 24GB
FP8 or INT8 weights, 18.6 GB.
Best for fine-tuning
1 GPU × NVIDIA A100 PCIe 80GB
LoRA on BF16 weights, 43.9 GB.
Budget pick
1 GPU × NVIDIA GeForce RTX 5080 16GB
Smallest card class that holds a 4-bit build, 9.3 GB.
FLUX.1 [schnell] VRAM requirements by precision
Model weights plus a 10 % margin for activations and runtime buffers, at one request.
| Precision | Weights | KV cache | Total for inference |
|---|---|---|---|
| BF16 / FP16 | 33.8 GB | Not applicable | 37.2 GB |
| FP8 | 16.9 GB | Not applicable | 18.6 GB |
| INT4 | 8.5 GB | Not applicable | 9.3 GB |
FLUX.1 [schnell] VRAM by task
Fine-tuning rows assume a short sequence, batch 1 and gradient checkpointing; longer sequences add activation memory.
| Task | GPU memory |
|---|---|
| QLoRA (4-bit base) | 11 GB |
| LoRA (BF16 base) | 43.9 GB |
| Full fine-tuning (BF16, AdamW) | 297.4 GB |
GPUs that fit FLUX.1 [schnell]
Number of cards whose combined memory holds each setup, assuming the model is split across them with tensor or pipeline parallelism. A dash means more than eight cards.
| GPU | Memory | BF16 inference (37.2 GB) | FP8 inference (18.6 GB) | INT4 inference (9.3 GB) | QLoRA (11 GB) | LoRA (43.9 GB) |
|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 5080 16GB | 16 GB | 4 GPUs | 2 GPUs | 1 GPU | 1 GPU | 4 GPUs |
| NVIDIA GeForce RTX 4090 24GB | 24 GB | 2 GPUs | 1 GPU | 1 GPU | 1 GPU | 2 GPUs |
| NVIDIA L4 24GB | 24 GB | 2 GPUs | 1 GPU | 1 GPU | 1 GPU | 2 GPUs |
| NVIDIA GeForce RTX 5090 32GB | 32 GB | 2 GPUs | 1 GPU | 1 GPU | 1 GPU | 2 GPUs |
| NVIDIA Tesla V100 32GB | 32 GB | 2 GPUs | 1 GPU | 1 GPU | 1 GPU | 2 GPUs |
| NVIDIA A100 PCIe 40GB | 40 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 2 GPUs |
| NVIDIA L40S 48GB | 48 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA Quadro RTX 8000 48GB | 48 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA RTX 6000 Ada 48GB | 48 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA RTX A6000 48GB | 48 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA A100 PCIe 80GB | 80 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA A100 SXM 80GB | 80 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA H100 PCIe 80GB | 80 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA H100 SXM 80GB | 80 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA RTX PRO 6000 Blackwell 96GB | 96 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA H200 141GB | 141 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA B200 180GB | 180 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA B300 288GB | 288 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA GB300 NVL | 288 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
What is FLUX.1 [schnell]?
FLUX.1 [schnell] is the fastest model of Black Forest Labs' FLUX.1 family: a 12-billion-parameter rectified flow transformer distilled to generate images from text in 1 to 4 steps, released under Apache 2.0 for personal, scientific and commercial use.
- Developer
- Black Forest Labs
- Released
- August 2024
- Parameters
- 16.9B
- License
- Apache 2.0
Same pipeline as FLUX.1 [dev]: the transformer, the T5-XXL (4.8B) and CLIP-L text encoders and a VAE, about 16.9 billion parameters. Offloading the T5 encoder to the CPU between prompts lowers the peak to about the transformer alone.
Source: FLUX.1 [schnell] model card (as of November 17, 2026).
Frequently asked questions
- How much VRAM does FLUX.1 [schnell] need?
- For inference, FLUX.1 [schnell] needs about 37.2 GB in BF16, 18.6 GB in FP8, 9.3 GB in INT4, including a 10 % runtime margin. Real usage depends on the inference engine and its settings.
- Can FLUX.1 [schnell] run on a single GPU?
- Yes. In BF16 it needs about 37.2 GB, which fits one NVIDIA L40S 48GB (48 GB).
- How much GPU memory does fine-tuning FLUX.1 [schnell] take?
- QLoRA (4-bit base): about 11 GB; LoRA (BF16 base): about 43.9 GB; Full fine-tuning (BF16, AdamW): about 297.4 GB. Longer training sequences and larger batches add activation memory on top.
- What is the best GPU for FLUX.1 [schnell]?
- Best for inference: 1 GPU × NVIDIA L4 24GB (FP8 or INT8 weights, 18.6 GB). Best for fine-tuning: 1 GPU × NVIDIA A100 PCIe 80GB (LoRA on BF16 weights, 43.9 GB). Budget pick: 1 GPU × NVIDIA GeForce RTX 5080 16GB (Smallest card class that holds a 4-bit build, 9.3 GB). Check live offers before booking, as availability changes.
- Can FLUX.1 [schnell] run on an RTX 4090?
- Yes. In FP8 it needs about 18.6 GB, which fits one RTX 4090 (24 GB). FP16 needs about 37.2 GB, 2 of these cards.
- Can FLUX.1 [schnell] run on an RTX 5090?
- Yes. In FP8 it needs about 18.6 GB, which fits one RTX 5090 (32 GB). FP16 needs about 37.2 GB, 2 of these cards.
Estimates use the formulas of the LLM VRAM calculator (decimal GB). Real usage depends on the inference engine, quantization format and memory fragmentation: test on the target GPU before a long rental.
Explore the live market
Every offer on the live market comes from a reviewed supplier and is sold by EverXCloud Inc.