Falcon3 models
4 Falcon3 models from TII on Hugging Face, from 1.7B to 7.5B parameters, published by TII in BF16, with contexts from 4K to 32K tokens. The smallest official model, Falcon3-1B-Base, needs about 3.7 GB of VRAM at its published precision; the cheapest live fit is RTX 3070 at $0.088/hr.
Part of the Falcon series · Previous generation: Falcon 1
Pick a size
One row per official Falcon3 size: the VRAM it needs at each precision, the cheapest GPU that holds it today, and the KV cache for a 32K-token context.
| Model | Parameters | Native VRAM | FP8 VRAM | INT4 VRAM | Live GPU fit (native) | Est. $/hr | KV cache at 32K |
|---|---|---|---|---|---|---|---|
| Falcon3-1B-Instruct | 1.7B | 3.7 GB | 1.9 GB | 0.9 GB | RTX 3070 | $0.088/hr | 2.3 GB |
| Falcon3-7B-Instruct | 7.5B | 16.7 GB | 8.3 GB | 4.2 GB | RTX A5000 | $0.176/hr | 3.5 GB |
VRAM is weights times a flat 1.2 overhead; the KV cache is a separate per-model figure at 16-bit, shown where the architecture is published. See the methodology.
Official models (4)
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| Falcon3-1B-Base | 1.7B | BF16 | 3.7 GB | RTX 3070 | $0.088/hr |
| Falcon3-1B-Instruct | 1.7B | BF16 | 3.7 GB | RTX 3070 | $0.088/hr |
| Falcon3-7B-Base | 7.5B | BF16 | 16.7 GB | RTX A5000 | $0.176/hr |
| Falcon3-7B-Instruct | 7.5B | BF16 | 16.7 GB | RTX A5000 | $0.176/hr |
VRAM is for the precision the model is published in: the weights times a flat 1.2 overhead, with the KV cache not included. See the methodology. The fit shown is the lowest-priced single GPU type that holds the model at that precision, or the lowest-priced multi-GPU set (up to 8) when none does.
Run and fine-tune Falcon3
- How to run Falcon3: engine support, chat template and context flags, with a serving command sized per model.
- How to fine-tune Falcon3: LoRA and QLoRA memory per model, and the license limits on derivatives.
More from TII
- Falcon 1 (4 models)
- Falcon-H1 (4 models)
- Falcon Mamba (2 models)