FLUX.1 Kontext
Image Available Full comparison ↗ bfl/flux-kontext · by Black Forest Labs
· Flow matching / rectified flow
Pricing — 1 offering(s)
[pro]
- $0.040 / image Current 2026-07-23 → present
[max]
- $0.080 / image Current 2026-07-23 → present
Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.
See how FLUX.1 Kontext fits into a cost-aware routing setup
See how →Capability profile
Operator guidance
Route here specifically for image-editing workloads (an existing image + a text instruction) rather than pure text-to-image generation — for the latter, the base FLUX.1/1.1 [pro] line (flux-1-1-pro.yaml) or FLUX.1 [dev]/[schnell] are the fit. Within Kontext, [pro] is the faster/cheaper default and [max] the higher-fidelity, pricier tier. Black Forest Labs' own documentation now recommends FLUX.2 over Kontext for new projects (superior quality, multi-reference support, up to 4MP output) — Kontext should be treated as the fallback/legacy editing option until a FLUX.2 ontology/registry entry exists.
Use cases
- Iterative, instruction-driven edits to an existing image (style changes, object edits, text replacement) rather than generating from scratch
- Character- and style-consistent edit series across multiple turns
- Existing integrations already built against FLUX.1 Kontext, per BFL's own migration guidance toward FLUX.2 for new projects
Limitations
- No independently published GenEval/T2I-CompBench score found for Kontext specifically — capability ratings above are vendor-stated, not third-party verified. An independent Elo does now exist for both priced tiers: Artificial Analysis Text to Image Arena (artificialanalysis.ai/image/leaderboard/text-to-image, confirmed 2026-08-19, SCO-462 sweep) ranks 'FLUX.1 Kontext [max]' #63 of ~147 models tracked (Elo 1141) and 'FLUX.1 Kontext [pro]' #79 (Elo 1109) — a different benchmark type than GenEval/T2I-CompBench, not a substitute for one
- Black Forest Labs' own documentation positions this as a previous-generation model and actively steers new projects to FLUX.2, which is not yet tracked anywhere in this registry
- Closed weights for [pro]/[max] — API-only, no self-hosting or fine-tuning for those tiers (a separate open-weight [dev] variant exists but is out of scope for this doc's pricing-linked entry)