MODEL · BLACK FOREST LABS · 4B (STEP-DISTILLED, ~4 INFERENCE STEPS) · 9B (PARENT FLOW MODEL)
FLUX.2 [klein] (4B + 9B)
FLUX distilled for fast inference. The 4B variant is Apache 2.0 — the first FLUX-quality image model you can actually ship in a commercial product.
License: 4B: Apache 2.0 · 9B: non-commercial · Context: Standard FLUX resolutions; sub-second on enterprise GPUs · Released: January 15, 2026
The decision in five lines
- The call
- Consider — for image
- Best for
- image
- Runs on
- 20 hardware picks fit (cheapest: Minisforum UM890 Pro · $463 barebone (no RAM / SSD); $991 with 32 GB / 1 TB — Minisforum US store, Sep 15 2026)
- Watch out
- Frontier prompt adherence — FLUX.2 [dev] still leads on complex compositions.
- Evidence
- Estimated
- 4B (step-distilled, ~4 inference steps) · 9B (parent flow model)
- PARAMETERS
- IMAGE GEN
- TYPE
- Standard
- CONTEXT
- ~13 GB (4B native) / ~29 GB (9B); FP8 variants halve that
- VRAM AT Q4
Where we recommend this
Every tier slot in the planner where this model is a top or alternate pick. Pulled live from planner.js — when the planner refreshes, this table stays current.
The call
FLUX distilled for fast inference. The 4B variant is Apache 2.0 — the first FLUX-quality image model you can actually ship in a commercial product.
When not to use: Frontier prompt adherence — FLUX.2 [dev] still leads on complex compositions. Klein is about speed + licensing, not quality ceiling. Also: 4B and 9B have OPPOSITE licenses — verify the string in the HF repo before committing to either.
Runner notes
ComfyUI or diffusers. 4B fits 8 GB VRAM with GGUF quant; 9B needs ~29 GB native (RTX 4090 and above) or ~15 GB at FP8. Apache 2.0 on 4B makes it the safe commercial default in the FLUX lineup.
Hardware that fits
Every hardware pick whose memory fits this model at the quant we recommend. Sorted cheapest-first — the top row is your best-value fit. Click through for the full buyer’s guide.
- Minisforum UM890 ProPerfect · 1.7× 32 GB DDR5 (shared) · $463 barebone (no RAM / SSD); $991 with 32 GB / 1 TB — Minisforum US store, Sep 15 2026
- RTX 5060 Ti 16 GBGood · 1.1× 16 GB · $700 (refurb / open box) – $970 (new)
- AMD Radeon RX 9070 XTGood · 1.1× 16 GB · $750–$870
- NVIDIA RTX 3090 (used, single)Perfect · 1.7× 24 GB · $950–$1,200
- NVIDIA RTX 5070 TiGood · 1.1× 16 GB · $1,150–$1,300 new; open-box / refurb from ~$1,020
- AMD Radeon RX 7900 XTXPerfect · 1.7× 24 GB · $1,300–$1,500 new (no refurb in stock Sep 15 2026)
- NVIDIA RTX 5080Good · 1.1× 16 GB · $1,430–$1,700
- MacBook Air M5 24 GBGood · 1.1× 24 GB unified · $1,499–$1,899
- Mac Mini M4 Pro 24 GBGood · 1.1× 24 GB unified · $1,699 (M5 Pro successor, 24 GB / 512 GB, pre-order; M4 Pro discontinued Aug 25 2026)
- Dual RTX 3090 (used)Perfect · 3.3× 48 GB · $1,800–$2,500 all-in
- NVIDIA RTX 4090Perfect · 1.7× 24 GB · $2,200–$2,800
- M5 Pro MacBook Pro 48 GBPerfect · 2.2× 48 GB unified · $2,999–$3,599
- Framework Desktop (Ryzen AI Max+ 395)Perfect · 5.9× 128 GB unified · $3,449 (128 GB config)
- NVIDIA RTX A6000 (48 GB, used)Perfect · 3.3× 48 GB ECC · $3,500–$4,500
- Mac Studio M4 Max 64 GBPerfect · 3.0× 64 GB unified · $3,799 (M5 Max successor, 64 GB / 1 TB, pre-order; M4 Max discontinued Aug 25 2026)
- NVIDIA DGX SparkPerfect · 5.9× 128 GB unified · $4,699
- M5 Max MacBook Pro 64 GBPerfect · 3.0× 64 GB unified · ~$5,199 (est.; June 25 2026 increase)
- Mac Studio M3 Ultra 96 GBPerfect · 4.4× 96 GB unified · $5,499 (M5 Ultra successor, 96 GB / 1 TB, pre-order; M3 Ultra discontinued Aug 25 2026)
- NVIDIA RTX 5090Perfect · 2.2× 32 GB · $6,450–$7,000 (new, in stock)
- Dual RTX 5090Perfect · 4.4× 64 GB (2×32) · $13,300–$14,000 all-in (two cards alone are ~$12,900 at the Sep 15 2026 Newegg floor)
Next step
Find-by-model — see what hardware runs this→