What GPU do I need to run poolside/Laguna-S-2.1-FP8?
117.6B parameters, published in F8_E4M3. View on Hugging Face
Set up Laguna-S-2.1-FP8Laguna-S-2.1-FP8 is published by poolside on Hugging Face, with 62,967 downloads and 27 likes to date. It's a LagunaForCausalLM model built for text-generation, published natively in F8_E4M3.
VRAM required & cheapest live GPU fit
Required VRAM = weight size at each precision, plus a fixed overhead for KV-cache, activations, and fragmentation. Full formula and assumptions: methodology.
| Precision | Weight size | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| FP8 (native) | 109.5 GB | 131.4 GB | RTX 4000 SFF Ada | 7 | $1.39/hr |
| INT4 (quantized) | 54.7 GB | 65.7 GB | A100 | 1 | $1.31/hr |
| cheaper alt. | RTX A5000 | 3 | $0.528/hr |
A GPU is only matched to a row if its hardware supports that precision, and the primary recommendation is always a single-GPU fit when one exists.
INT4 requires a quantized checkpoint actually published for this model, check its Hugging Face page before relying on this row.
Cheapest way to run Laguna-S-2.1-FP8 at its published (F8_E4M3) precision: 7× RTX 4000 SFF Ada, at $0.198/hr per GPU ($1.39/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
Laguna-S-2.1-FP8: common questions
Can Laguna-S-2.1-FP8 run on a single GPU?
No. At FP8 (native) it needs 131.4 GB of VRAM, more than a single 32 GB desktop card holds. The cheapest capable card in the live feed is a 20.0 GB RTX 4000 SFF Ada, and it takes 7 of them.
Is Laguna-S-2.1-FP8 already quantized?
Yes. It is published in FP8, one byte per parameter, so the 131.4 GB figure is already a quantized footprint rather than a full-precision one. Only INT4 goes below it, at 65.7 GB. A GPU without FP8 tensor cores cannot run it as published, which is why cards here are matched on precision support and not on VRAM alone.
How many GPUs do I need to run Laguna-S-2.1-FP8?
7 at FP8 (native). It needs 131.4 GB of VRAM and the cheapest capable live offer is a 20.0 GB RTX 4000 SFF Ada, so 7 of them come to $1.39/hr in total.
Does quantizing Laguna-S-2.1-FP8 lower the GPU bill?
Yes. At FP8 (native) the cheapest live fit is 7 RTX 4000 SFF Ada cards at $1.39/hr. At INT4 (quantized) it drops to one A100 at $1.31/hr, provided a quantized checkpoint exists for it.
Weight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
More Laguna models
- Laguna-S-2.1 (117.6B, BF16)
- Laguna-XS-2.1 (33.4B, BF16)
- Laguna-XS.2 (33.4B, BF16)
Related reading: H100 pricing and specs, The best GPUs for AI, ranked, and Best GPU for LLM inference.
