Meta · 3D · SAM License
SAM 3D Objects
Needs 16 GB at BF16 / FP16. The smallest real configuration today is 1× T4 at Vast.ai, $0.04 per hour, about $29 a month at 730 hours.
Parameters—not published
Category3Dsized on weights + headroom
Checkpoint13 GBas released, as published
LicenseSAM Licensecheck before commercial use
ReleasedNov 19, 2025documented date
Smallest configuration · BF16 / FP16
1× T416 GB VRAM · 98% used
- Per hour
- $0.04
- Per month
- $29
- Cheapest at
Vast.ai
What is documented about SAM 3D Objects
news checked 2026-10-27Single image + mask to 3D shape, texture and layout; SAM 3D Body is the human-mesh sibling. Not in the catalogue.
- https://huggingface.co/facebook/sam-3d-objects
- https://huggingface.co/api/models/facebook/sam-3d-objects?blobs=true
How much VRAM it needs
Weights plus 20% runtime overhead. For 3d models, activations also depend on resolution, clip length and batch size; the per-GPU estimates below give the practical configuration.
| Precision | Weights | Weights + headroom |
|---|---|---|
| BF16 / FP16estimated | 13 GB | 16 GB |
Cross-checked on Hugging Face Oct 28, 2026: repository read. Method: weights estimated from the parameter count (2, 1 and 0.6 bytes per parameter; the published checkpoint size at its own precision) x 1.2 runtime headroom; short context.
What it costs to run, GPU by GPU
The fewest GPUs of one type whose total memory fits the model at short context, priced from the cheapest eligible on-demand offer for exactly that count (45-day window, newest reading per provider), with the next providers beside it. Monthly = 730 hours.
| GPU | Smallest config | VRAM used | $ / hour | $ / month | Cheapest at |
|---|---|---|---|---|---|
| T4Best fit | 1× 16 GB16 GB total | 98% tight | $0.04 | $29 | |
| Tesla P40 | 1× 24 GB24 GB total | 66% | $0.04 | $30 | |
| Tesla P100 | 1× 16 GB16 GB total | 98% tight | $0.04 | $33 | |
| RTX A4000 | 1× 16 GB16 GB total | 98% tight | $0.06 | $46 | |
| Tesla V100 16GB | 1× 16 GB16 GB total | 98% tight | $0.06 | $46 | |
| RTX A2000 | 2× 12 GB24 GB total | 66% | $0.07 | $50 | |
| RTX 3060 | 2× 12 GB24 GB total | 66% | $0.07 | $50 | |
| Tesla V100 32GB | 1× 32 GB32 GB total | 49% | $0.07 | $50 | |
| RTX PRO 2000 | 1× 16 GB16 GB total | 98% tight | $0.07 | $51 | |
| Tesla M40 | 1× 24 GB24 GB total | 66% | $0.07 | $54 | |
| Titan Xp | 2× 12 GB24 GB total | 66% | $0.08 | $60 | |
| RTX 5060 Ti | 1× 16 GB16 GB total | 98% tight | $0.09 | $64 | |
| RTX 3090 | 1× 24 GB24 GB total | 66% | $0.09 | $66 | Akash NetworkTrust 53 · read Nov 1 · receipt · also Vast.ai $0.10, Clore.ai $0.11 |
| GTX 1070 | 2× 8 GB16 GB total | 98% tight | $0.10 | $70 | |
| GTX 1070 Ti | 2× 8 GB16 GB total | 98% tight | $0.10 | $76 | |
| L4 | 1× 24 GB24 GB total | 66% | $0.10 | $76 | |
| GTX 1080 | 2× 8 GB16 GB total | 98% tight | $0.11 | $79 | |
| RTX 2070 | 2× 8 GB16 GB total | 98% tight | $0.11 | $79 | |
| Quadro P2000 | 4× 5 GB20 GB total | 79% | $0.11 | $80 | |
| RTX 3060 Ti | 2× 8 GB16 GB total | 98% tight | $0.11 | $81 | |
| RTX A4500 | 1× 20 GB20 GB total | 79% | $0.11 | $82 | |
| RTX A5000 | 1× 24 GB24 GB total | 66% | $0.12 | $87 | |
| RTX 2070 Super | 2× 8 GB16 GB total | 98% tight | $0.12 | $89 | |
| RTX 4060 | 2× 8 GB16 GB total | 98% tight | $0.12 | $89 | |
| The 24 cheapest of 98 GPU types that fit. Every GPU, any precision and context, in the finder | |||||
