Stable Diffusion 3.5 Large.
Stability AI's latest open-weight image-gen — 8.1B params, MMDiT architecture.
1× Nvidia GeForce RTX 2060.
Most-aggressive quantisation we have a working recommendation for. Lower precision = less VRAM = cheaper hardware, at a small accuracy cost.
Cheapest hosted endpoints.
| Provider | Access | $/M in | $/M out | |
|---|---|---|---|---|
| Stability AI API | api direct | — | — | |
| Replicate | hosted inference | — | — |
Smaller models distilled from Stable Diffusion 3.5 Large.
Lightweight student models trained to mimic Stable Diffusion 3.5 Large's outputs.
Variants in the Stable Diffusion family.
Stable Diffusion 3.5 Medium
2.5B SD 3.5 — fits on 12 GB consumer GPUs.
Stable Diffusion XL
Workhorse open-weight image-gen — 3.5B params, runs anywhere.
Stable Diffusion 1.5
The original viral image-gen model — still searched heavily.
Frequently asked.
How do I run Stable Diffusion 3.5 Large?
Where can I access Stable Diffusion 3.5 Large?
How much does it cost to run Stable Diffusion 3.5 Large?
Is Stable Diffusion 3.5 Large open-source or proprietary?
Cheapest hardware per quantisation.
Each row is one quantisation tier (the same weights compressed differently). Lower precision → lower VRAM → cheaper hardware, at the cost of small accuracy loss. $/hr refreshed hourly from each provider's API.
| Quantisation | Cheapest GPU config | Total VRAM | Live $/hr | tokens/sec | |
|---|---|---|---|---|---|
|
FP16
FP16 — half precision (default)
|
24 GB | — | — | Compare → | |
|
FP8
FP8 — 8-bit float (Hopper / Blackwell)
|
FP8 cuts VRAM by half.
|
24 GB | — | — | Compare → |
|
INT4
INT4 — 4-bit integer (~4× VRAM saving)
|
6 GB | — | — | Compare → |
What it costs per month across providers.
Estimate your monthly bill for Stable Diffusion 3.5 Large across every host that publishes per-token pricing. Slide your token volumes; the chart + table re-rank cheapest-first.
No priced API access rows on file for Stable Diffusion 3.5 Large yet.
Rent the GPU instead of paying per token.
For an open-weights model like Stable Diffusion 3.5 Large, you can rent a GPU and serve inference yourself. The math: cheapest GPU rental × 730 hours/month + your electricity rate × power draw.
Assumes the GPU runs 24/7 at ~85% utilisation. If your traffic is bursty, you'll pay less for the API and probably more for the GPU (idle hours still cost rental). The breakeven analysis lives on the Self-host vs API breakeven tool.
About Stable Diffusion 3.5 Large.
Stable Diffusion 3.5 Large is Stability AI's late-2024 flagship open-weight image-gen model — Multimodal Diffusion Transformer (MMDiT) architecture with 8.1B parameters. Improves prompt adherence and text rendering significantly vs SDXL. Community License: free for research + ≤$1M ARR commercial use; enterprise license required above that.