Riverflow 2.0 Pro fits creative agencies and e-commerce teams that need reliable, legible in-image text for packaging and ad work, not casual or open-ended art generation. Its 8,192-token prompt window leaves room for detailed multi-element briefs, and the agentic self-correction pass catches text and layout errors before output, replacing a manual retouching step most competing image models still require.
Riverflow 2.0 Pro is Sourceful's agentic image generation and editing model, the first non-major-lab release to rank #1 on Artificial Analysis in February 2026 across both image categories at once. Its self-correction pipeline and font control produce commercial-grade, legible text inside generated images, a persistent weakness for standard diffusion models.
Provider: Sourceful · Family: Riverflow 2.0
Context window: 8,192 tokens
Input modalities: text, image · Output: image
About Riverflow 2.0 Pro
Riverflow 2.0 Pro is an agentic image generation and editing model built by Sourceful, a Manchester, UK company founded in 2020. It launched in February 2026 as the flagship tier of the Riverflow 2.0 suite, extending Sourceful's October 2025 Riverflow 1 model from image editing only into combined text-to-image and editing under one agentic reasoning pipeline. After an initial generation pass, the model's reasoning component checks prompt fidelity and spatial accuracy before returning the final image, an approach built for commercial brand design where following complex instructions matters more than raw speed.
On Artificial Analysis's All Listings leaderboards, Riverflow 2.0 ranked #1 in the Text to Image and Image Editing categories simultaneously, ahead of GPT Image 1.5 (high), Nano Banana Pro (Gemini 3 Pro Image), and FLUX.2 [max]. Artificial Analysis runs the most widely cited third-party image-model leaderboard, and this was the first time a challenger outside the major labs topped both image-quality categories at once.
Riverflow 2.0 Pro accepts prompts up to 8,192 tokens, takes reference images for image-to-image editing, and outputs up to 4K resolution with aspect ratios as wide as 21:9 for cinematic compositions. The Pro tier trades speed for output quality; the Fast and Standard Preview variants in the same suite prioritize lower latency instead.
Its most distinctive capability is text rendering: users can name a typeface directly in the prompt, and the pipeline checks that on-image text is legible, positioned correctly, and matches the request. That addresses a longstanding diffusion-model weakness, and is why the model is competitive for packaging labels, ad copy, and branded collateral.
There is no direct Sourceful API; access runs only through OpenRouter (model ID sourceful/riverflow-v2-pro), Runware, Replicate, WaveSpeed AI, or an enterprise agreement at riverflow.ai. Weights are closed and not downloadable, so self-hosting is not an option.
Safety checking is configurable: users can run a single-pass content check before generation, disable it for trusted commercial pipelines, or scan only user-submitted prompts. Sourceful has not released a system card or independent safety evaluation, and commercial use is governed entirely by its API terms of service.
Riverflow 2.5 Pro, released June 18, 2026, extends 4K output to the standard tier and adds custom scoring for brand style consistency; it topped all three Design Arena categories at once, with an Elo of 1408 in Image, 1469 in Graphic Design, and 1377 in Image Editing. Teams starting new projects should evaluate 2.5 Pro first; 2.0 Pro remains a reasonable choice for teams already built on its API.
Pricing
$0.15 per generated image at 1K or 2K resolution; $0.33 per 4K image. Pricing is per output image, not per token. Provider pricing on OpenRouter, Runware, and Replicate may vary slightly.
Key Features
- Agentic Self-Correction Pipeline: Runs a second internal pass that checks the first generation against the prompt for fidelity and spatial accuracy before returning any output.
- Font-Accurate Text Rendering: Lets you specify a font by name in the prompt; the self-correction pass confirms the on-image text is spelled right and sits where the prompt asked before returning the final image.
- 4K Output, Wide Aspect Ratios: Produces images at up to 4K with a 21:9 maximum aspect ratio, enough width for billboard and cinematic-format layouts.
- Reference-Based Image Editing: Accepts a reference image plus a natural-language edit instruction, and verifies the edit matches the request before finalizing the output.
- Multi-Platform API Access: Reachable through OpenRouter, Runware, Replicate, and WaveSpeed AI with standard API-key auth, so there is no proprietary SDK to learn.
Pros
- Ranked #1 on Artificial Analysis across both text-to-image and image editing at once, a benchmark result no single non-major-lab model had hit before it launched.
- Agentic self-correction catches prompt-fidelity and spatial errors before the image is returned, which matters most on complex multi-element product or packaging scenes.
- Font control renders legible, correctly positioned on-image text, the one diffusion-model weakness most competitors still struggle with.
Cons
- Closed weights and API-only access rule out self-hosting, fine-tuning, or air-gapped deployment.
- Per-image pricing gets expensive fast at high volume compared to self-hosted open-weights models running on owned GPUs.
- Sourceful's own Riverflow 2.5 Pro already beats it on Design Arena scores, so check that newer tier before committing to 2.0 Pro.
Benchmarks
A benchmark is an exam, not the job. Scores transfer unevenly between tasks, so weigh the one closest to your workload and read every figure with its source.
Frequently Asked Questions
What are Riverflow 2.0 Pro's pricing plans in 2026?
Riverflow 2.0 Pro bills per generated image, not per token or subscription: $0.15 for a 1K or 2K image, and $0.33 for a 4K image. A batch of 100 product mockups at 2K runs about $15, and pricing is roughly the same whether you route through OpenRouter, Runware, or Replicate. There is no annual or volume discount published as of June 2026.
Is Riverflow 2.0 Pro free to use?
No. Riverflow 2.0 Pro has no free tier or trial credits; every generated image is billed from the first request. There is no subscription or minimum spend either, so teams can evaluate output quality by paying for a handful of test images through OpenRouter, Runware, or Replicate rather than committing to a plan.
What are the best alternatives to Riverflow 2.0 Pro?
FLUX.1 [dev] and SD3.5 are the main open-weights alternatives for teams that want to self-host on their own GPUs instead of paying per image. FLUX.2 [max] is the closest proprietary rival on photorealistic natural imagery, though Riverflow 2.0 Pro currently ranks ahead of it on the Artificial Analysis leaderboards for both text-to-image and image editing. GPT Image 2, OpenAI's own 2026 image flagship, is another proprietary option worth comparing on token-based billing.
Riverflow 2.0 Pro or FLUX.2 [max]: which should you pick?
Riverflow 2.0 Pro currently tops Artificial Analysis's leaderboards in text-to-image and image editing at once, and its agentic self-correction pipeline targets legible, correctly positioned on-image text, which FLUX.2 [max] does not specialize in. FLUX.2 [max] is open-weights and can be self-hosted, and it remains the stronger pick for photorealistic natural imagery. Choose Riverflow 2.0 Pro for commercial brand and packaging work with text; choose FLUX.2 [max] if you need to run inference on your own GPUs.
What does it take to start using Riverflow 2.0 Pro?
There's no self-serve dashboard on Sourceful's own site: grab an API key from OpenRouter, Runware, Replicate, or WaveSpeed AI and call the model by its ID, sourceful/riverflow-v2-pro on OpenRouter. Write a text prompt with the subject first and style modifiers after, since the agentic pipeline anchors most strongly to whatever comes first. For brand or packaging work needing on-image text, name the typeface directly in the prompt to activate font control.
Top Alternatives
- FLUX.2 [max]: Pick FLUX.2 [max] for photorealistic natural imagery and exact color matching; pick Riverflow 2.0 Pro for legible in-image text and a top Artificial Analysis rank.
- GPT Image 2: Pick GPT Image 2 for token-based billing and tighter OpenAI ecosystem integration; pick Riverflow 2.0 Pro for its agentic self-correction and brand-design text accuracy.
- Gemini 3 Pro Image: Pick Gemini 3 Pro Image for Search-grounded factual accuracy; pick Riverflow 2.0 Pro for its top Artificial Analysis rank across both text-to-image and editing.
- MAI-Image-2.5: Pick MAI-Image-2.5 if you are already on Microsoft's stack; pick Riverflow 2.0 Pro for the higher Artificial Analysis rank in both categories.