This page covers FLUX.1 [pro] text-to-image models that are still callable on APIYI. The latest FLUX.2 generation and the editing-focused FLUX.1 Kontext are documented in the FLUX Overview.
Version Snapshot
Per-Version Specs
flux-pro-1.1-ultra
- Released: 2024-11 (UTC+0)
- APIYI Price: $0.0500 (official $0.06, saves 17%)
- Max output: ~4MP (highest in FLUX.1 series)
- Key features: Ultra-resolution, optional raw mode (closer to candid photography)
- Known limits: Single reference only, no hex color control, no grounding search
- Official docs:
docs.bfl.ai/flux_models/flux_1_1_pro_ultra_raw
flux-pro-1.1
- Released: 2024-10 (UTC+0)
- APIYI Price: $0.0350 (official $0.04, saves 12.5%)
- Max output: ~1.6MP (e.g., 1024×1536)
- Key features: Improved quality and prompt adherence over 1.0, industry baseline
- Known limits: Single reference, short prompts (no 32K)
- Official docs:
docs.bfl.ai/flux_models/flux_1_1_pro
flux-pro
- Released: 2024-08 (UTC+0)
- APIYI Price: $0.0400 (official $0.04, same)
- Max output: ~1.6MP
- Key features: BFL’s first commercial pro, baseline text-to-image
- Known limits: Quality and adherence below 1.1 — no reason for new projects
flux-dev
- APIYI Price: $0.0200
- Max output: ~1MP
- Key features: Open-weights variant (FLUX.1 [dev], non-commercial license), self-host on your own GPU
- Known limits: Quality below the [pro] tier, mainly for research, prototyping, local inference validation
- Official weights:
huggingface.co/black-forest-labs/FLUX.1-dev
Migration Guidance
1
Assess differences
FLUX.2 fully supersedes FLUX.1 [pro]: 4MP output (vs 1.6MP), up to 8 multi-references (vs 1), 32K-token prompts (vs short), native hex color control, typography specialist tier. Pricing within 1MP is similar or lower.
2
Run side-by-side
Generate the same prompt set on
flux-pro-1.1 and flux-2-pro for a week. Compare text legibility, multi-object consistency, brand color fidelity. FLUX.2 [pro] wins across the board in most cases.3
Roll out gradually
Cut 10% of traffic to
flux-2-pro first, observe quality and cost for a week, then ramp to 100%. Costs are roughly equivalent at 1MP and noticeably cheaper at 4MP.4
Handle parameter differences
Most parameters carry over, but note:
- FLUX.1 [pro] takes a single reference image; FLUX.2 supports multiple references (JSON fields
input_image~input_image_8, up to 8) - FLUX.1 doesn’t support
prompt_upsampling; FLUX.2 [pro/max/flex] does - Some legacy aspect-ratio identifiers (e.g.,
aspect_ratio) are replaced bywidth/heightor thesizestring in FLUX.2
Legacy Call Example
Cost Comparison
By typical volume (flat per-image price):How to choose: For new projects, prefer FLUX.2 (
flux-2-pro for general use, flux-2-max for flagship). Keep FLUX.1 [pro] running only when legacy integrations are deeply wired and regression testing is impractical. flux-dev remains a viable low-cost dev/test option.Related Docs
- FLUX Overview — full model matrix and selection
- Text-to-Image Playground — works for FLUX.2 + FLUX.1
- Image Editing Playground — multi-reference fusion + edit