FLUX.2 klein 4B vs FLUX.1 dev vs Z-Image-Turbo Benchmarked Across Five GPUs From 12GB to 96GB
strata2signal · reddit · 2026-09-26
A small home studio that paints art for its local-only AI RPG published a Black Forest Labs history and license guide, plus a unified benchmark of three models on five GPUs.
Licenses: schnell and klein 4B are Apache 2.0; dev and klein 9B are non-commercial.
Benchmarks (same pinned ComfyUI stack, 18 renders, cards: 3080 Ti 12GB / 5080 16GB / laptop 5090 24GB / 3090 Ti 24GB / RTX PRO 6000 96GB):
- Seconds per 1024×1024 image at 8 steps (96GB→12GB): klein 4B 1.506→5.725; Z-Image-Turbo 1.993→8.604; FLUX.1 dev 2.758→10.857.
- Only klein 4B fit fully in 12GB; the other two paged in. Everything sat whole from 16GB up.
- klein 4B images per 10 minutes: RTX PRO 6000 252, 5080 164, laptop 132, 3090 Ti 117, 3080 Ti 101.
- Energy: the laptop (150W) and RTX PRO 6000 (300W) spent nearly identical energy per image (663 vs 667 J).
The studio now uses klein 4B daily across a print lab, a public art board and a browser drum machine.
More from Multimodal
- Comment: China leads the AI video revolution while the West argued about AI art — TinfoilTricorn · 2026-11-11
- "QT your Geisha" meme takes over X with stunning porcelain-style AI portraits — LudovicCreator · 2026-11-11
- MiniMax H3-Image early access test: 4 images, one coherent cyberpunk world — Aiden_Tech_Ai · 2026-11-11
- MiniMax H3 Image generates a GTA6-style hurricane scene from a single prompt — socialwithaayan · 2026-11-11
- Dragon Ball reimagined as a 1971 movie: another hit AI retro trailer — techhalla · 2026-11-11
- ReCaVSR lands in ComfyUI: one-step Wan2.2 video upscaler hits 4K at 4.65 fps on an RTX 4090 — LatentSpacer · 2026-11-11