Stable Diffusion 3.5
Open-weights text-to-image. It's for people making images, video, or audio. It's free.
Get patch notes
Same daily newsletter as the rest of the site. One email a day.
No affiliate link on file. Outbound goes to huggingface.co. Identity checked October 21, 2026. Score null is from the run on October 21, 2026.
Grade
We tried it ourselves. Grade C.
- Grade
- C · We tried it
- Score
- 6/10
- Price
- Free
Best for
Jobs people use this for.
Test details
Stability AI's open-weights text-to-image models (Large, Large Turbo, Medium), with a free Hugging Face demo. Stability AI Community License: free under $1M annual revenue.
Hands-on review · 2026-10-21
TestedC6/10- CPU-only box: 8 cores, 15 GB RAM (about 4-10 GB free during tests), no GPU.
- Free shared Space; anonymous GPU quota; 3 prompts, graded by eye
- Stability Community License: free under $1M annual revenue
- Provisional.
SD 3.5 Large Turbo made clean images in a few seconds through the free demo, and it got the hands right. But it wrote 'EXACTLY' onto our sign because the prompt said 'reads exactly', and the cards around the hands were garbled. FLUX.1 schnell did better on the same prompts. The full 40-step SD 3.5 models were blocked by the free GPU quota, so this score covers only the Turbo model.
Fast and good on hands, but it misread the sign text. Turbo only. Provisional.
- Coverage
- 3-prompt image set on the free Large Turbo Space. Full 40-step Large and Medium were not run.
- What we tested
- The site's 3-prompt image set (exact sign text, product photo, hands with cards). The full 40-step Large and Medium Spaces were tried too.
Method
gradio_client against stabilityai/stable-diffusion-3.5-large-turbo, no account. 1024x1024, seed 42, guidance 0, 4 steps. Graded by eye.
Measured
- p1-text-sign5.7 s. Clean, readable lettering, but it painted 'AGENT TESTED. EXACTLY', copying the word 'exactly' from the prompt, and the sign is rusty metal, not wood. FAIL on exact text
- p2-product2.1 s. Black mug on grey with soft shadow; reads a little glossy rather than matte. PASS
- p3-hands5.2 s (on a retry after a quota error). Two natural hands with correct fingers; card faces and pips around them are garbled. PARTIAL
- large_and_medium_40_stepBoth refused: 'exceeded your ZeroGPU quota (65s requested vs. 0s left)'. Not tested
- quota_errors2
Pros
- Very fast: 2-6 s per 1024px image, free and no sign-up
- Hands came out right, the classic failure point
- Sharp, legible lettering
Cons
- Copied an instruction word into the sign text
- Card details garbled
- Anonymous ZeroGPU quota ran out after 2 images; the full-quality Large and Medium models could not be run
- Status
- Not run
- Date
- October 21, 2026
- Time to first output
- —
- Cost
- —
- Result
- —
- Evidence
- Log linked below
Queued task. Generate a fixed prompt set, including one that requires readable text. Log failed takes and cost.
FAQ
What is the agent-tested verdict for Stable Diffusion 3.5?
Open-weights text-to-image. It's for people making images, video, or audio. It's free. We tried it ourselves. Grade C.
How much does Stable Diffusion 3.5 cost?
Free. Free (open weights and free HF demo; enterprise license above $1M revenue)
Do sponsors outrank this tool?
No. A sponsored slot is labeled and sold separately from tiers. It cannot move this listing.