FLUX.3 Image represents Black Forest Labs's newest flagship image generation model, launched on October 1, 2026, and immediately listed on OpenRouter for API access. The model performs two core workflows: generating images from text prompts and editing images using multi-reference inputs. Unlike traditional text-to-image approaches, the editing workflow accepts up to 10 reference images alongside a text prompt, enabling targeted modifications while preserving subject consistency.
The image shows the official FLUX.3 Image model listing on OpenRouter, which provides API access for Black Forest Labs' newest flagship image generation model. This visualization represents the core technical information discussed in the article, including resolution tiers, pricing structure, and performance metrics.
Output resolution is configurable across fixed tiers—768, 1K, 1.5K, 2K, and 4K—with aspect ratios including 21:9, 2:1, 16:9, 3:2, 7:5, and 4:3. This tiered structure allows users to balance quality requirements against cost, with pricing calculated as a flat rate per image that scales directly with the chosen resolution tier. At the lowest setting, generation starts from $0.0205 per image, rising to $0.3035 at the maximum 4K tier.
The pricing differs from Black Forest Labs' FLUX.2 family, which charges per megapixel rather than per image. The shift to flat-rate pricing makes costs predictable when resolution is fixed in advance, whereas the per-megapixel models could prove cheaper for small outputs. OpenRouter explains that caching and volume discounts typically reduce the amount customers actually pay below the listed rate.
Performance metrics from the OpenRouter listing show a median (P50) end-to-end latency of 34.85 seconds, with an uptime of 99.31% and availability of 98.61% over the preceding 72 hours. Because the model is served by a single provider with OpenRouter routing requests directly, these figures reflect that provider's responsiveness rather than a distributed load-balanced fleet.
The listing includes benchmark rankings showing where FLUX.3 Image places among evaluated models on standardized prompts, though the specific scores are not included in the provided content. Public applications already routing substantial traffic through the model include Weekend Willi Audience Classifier (1.07 million tokens), Cartoon Studio Image Engine (234K tokens), zapimage, RoomLab.app, and Paintme atmosphere photos, suggesting practical utility across classification, cartoon generation, and interior-design-type imagery.
Several questions raised in the Hacker News discussion cannot be answered from the supplied sources. The primary listing does not describe spatial or element-placement controls beyond the multi-reference editing mechanism, and there is no information about frame-by-frame sprite generation. The quality of the examples shown on the listing page is not evaluated by the source material. Whether FLUX.3 Image will be released as open weights, as was done with FLUX.2 Klein, is not stated.