NVIDIA · Image generation · 64.6B parameters · other
Cosmos3 Super Text2Image
Needs 78 GB at FP8 / INT8. The smallest real configuration today is 4× RTX 3090 at Clore.ai, $0.25 per hour, about $183 a month at 730 hours.
Parameters64.6Bdense
CategoryImage generationsized on weights + headroom
Checkpoint129 GBBF16, as published
Licenseothercheck before commercial use
Released—documented date
Smallest configuration · FP8 / INT8
4× RTX 309096 GB VRAM · 81% used
- Per hour
- $0.25
- Per month
- $183
- Cheapest at
Clore.ai
Memory · 3 precisionsHow much VRAM it needs
Weights plus 20% runtime overhead. For image generation models, activations also depend on resolution, clip length and batch size; the per-GPU estimates below give the practical configuration.
| Precision | Weights | Weights + headroom |
|---|---|---|
| BF16published checkpoint | 129 GB | 156 GB |
| FP8 / INT8estimated | 65 GB | 78 GB |
| INT4 / MXFP4estimated | 39 GB | 47 GB |
Cross-checked on Hugging Face Oct 28, 2026: 64.6B parameters in the safetensors headers, 132.5 GB of weight files. Method: weights as published at the native precision, verified on Hugging Face 2026-10-28; other precisions estimated at 1 (8-bit) and 0.6 (4-bit) bytes per parameter; × 1.2 runtime headroom; KV cache not added (architecture not readable).
Smallest configuration per GPU · 109 GPUs · cheapest eligible on-demand offer per GPU count, every provider on fileWhat 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 |
|---|---|---|---|---|---|
| RTX 3090Best fit | 4× 24 GB96 GB total | 81% | $0.25 | $183 | Clore.aiTrust 49 · read Nov 2 · receipt · also Vast.ai $0.56, Edgevana $0.92 |
| Tesla V100 32GB | 4× 32 GB128 GB total | 61% | $0.27 | $196 |  Vast.aiTrust 78 · read Nov 2 · receipt · also Cyfuture AI $2.22, gcube $6.44 |
| RTX A6000 | 2× 48 GB96 GB total | 81% | $0.38 | $274 | Clore.aiTrust 49 · read Nov 2 · receipt · also Lium $0.84, Spheron $0.89 |
| Tesla P100 | 8× 16 GB128 GB total | 61% | $0.38 | $274 | Vast.aiTrust 78 · read Nov 2 · receipt |
| RTX 3060 | 8× 12 GB96 GB total | 81% | $0.43 | $312 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.64, AtmosCompute $1.01 |
| A800 80GB PCIe | 1× 80 GB80 GB total | 97% tight | $0.47 | $342 | Vast.aiTrust 78 · read Nov 2 · receipt |
| Quadro RTX 8000 | 2× 48 GB96 GB total | 81% | $0.48 | $354 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.56 |
| L40 | 2× 48 GB96 GB total | 81% | $0.54 | $391 | Vast.aiTrust 78 · read Nov 2 · receipt · also io.net $1.32, GPU.ai $1.56 |
| RTX 2080 Ti | 8× 11 GB88 GB total | 88% | $0.59 | $432 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.68 |
| A100 80GB | 1× 80 GB80 GB total | 97% tight | $0.60 | $440 | Vast.aiTrust 78 · read Nov 2 · receipt · also packet.ai $0.89, GPUBank $0.95 |
| RTX PRO 6000 | 1× 96 GB96 GB total | 81% | $0.66 | $482 |  packet.ainot scored · read Nov 2 · receipt · also Runpod $0.69, HyperAI $0.80 |
| Titan RTX | 4× 24 GB96 GB total | 81% | $0.70 | $508 | Vast.aiTrust 78 · read Nov 2 · receipt |
| Tesla V100 16GB | 8× 16 GB128 GB total | 61% | $0.75 | $547 | Vast.aiTrust 78 · read Nov 2 · receipt · also CloudRift $2.32, Lambda $6.32 |
| L40S | 2× 48 GB96 GB total | 81% | $0.76 | $555 |  LiumTrust 40 · read Nov 2 · receipt · also Vast.ai $1.07, CloudRift $1.26 |
| A100 40GB | 2× 40 GB80 GB total | 97% tight | $0.77 | $566 | Vast.aiTrust 78 · read Nov 2 · receipt · also iwinv $1.83, AtmosCompute $1.93 |
| RTX A5000 | 4× 24 GB96 GB total | 81% | $0.80 | $586 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $1.05 |
| RTX 5060 Ti | 8× 16 GB128 GB total | 61% | $0.82 | $596 | Clore.aiTrust 49 · read Nov 2 · receipt · also Vast.ai $0.86, Edgevana $0.98 |
| RTX PRO 4000 | 4× 24 GB96 GB total | 81% | $0.83 | $608 | Clore.aiTrust 49 · read Nov 2 · receipt · also Vast.ai $1.55, Edgevana $1.78 |
| RTX 4070 Super | 8× 12 GB96 GB total | 81% | $0.85 | $617 | Vast.aiTrust 78 · read Nov 2 · receipt · also AtmosCompute $1.36 |
| A40 | 2× 48 GB96 GB total | 81% | $0.94 | $683 | Vast.aiTrust 78 · read Nov 2 · receipt · also Exoscale $2.09 |
| RTX PRO 5000 | 2× 48 GB96 GB total | 81% | $0.96 | $702 | Vast.aiTrust 78 · read Nov 2 · receipt · also Lium $1.70, AtmosCompute $2.04 |
| A10 | 4× 24 GB96 GB total | 81% | $0.96 | $703 | Vast.aiTrust 78 · read Nov 2 · receipt · also Volcano Engine $6.39 |
| RTX A4000 | 8× 16 GB128 GB total | 61% | $0.96 | $704 | Vast.aiTrust 78 · read Nov 2 · receipt |
| RTX 6000 Ada | 2× 48 GB96 GB total | 81% | $0.99 | $722 | Vast.aiTrust 78 · read Nov 2 · receipt · also GPU.ai $1.30, io.net $1.37 |
| The 24 cheapest of 61 GPU types that fit. Every GPU, any precision and context, in the finder | |||||
Throughput estimates · Oct 4, 2026Estimated speed, in s per image
Image generation. Speed estimated from memory bandwidth and model size, never measured. The cost per 1,000 images is an estimate at the prices read on Oct 4, 2026; it is not recomputed here.
| GPU | Cards | Precision | s per image | per 1,000 images · estimate |
|---|---|---|---|---|
| B200 | 1 | FP8 | 8.7 | $10.00 |
| H200 | 1 | FP8 | 16.3 | $11.30 |
| MI300X | 1 | FP8 | 25.0 | $11.40 |
cards, precision and speed estimated from memory bandwidth and model size; cost recomputed by the site from verified prices · cheapest per output at the prices of Oct 4, 2026: B200 SXMWhy estimates, and their limits →
Cosmos3 Super Text2Image · related