OpenRouter Image API
OpenRouter's dedicated Image API unifies access to 30+ image-generation models from 8 providers and exposes typed model capability discovery so agent code can validate image requests before calling a provider. Source: OpenRouter Blog, https://openrouter.ai/blog/announcements/image-api/, 2026-06-23
Routing Summary
Use OpenRouter Image API when an agent needs to choose an image model from typed capability metadata instead of a hardcoded provider list. Pair with Browser Testing Skills for product implementation, Frontend and Design Skills for design-artifact workflows, and ChatGPT Images 2.0 (gpt-image-2) when comparing against OpenAI's native image surface.
Agent Value
The /api/v1/images/models endpoint returns typed capability descriptors for every model. OpenRouter's thread calls this out as especially useful for coding agents: hand the agent the response and it can pick a model, validate inputs, and generate images without trial-and-error. Source: X/@OpenRouter thread, 2026-06-24
Normalized request shape covers resolution, aspect ratio, quality, output format, transparency, input references, streaming, and provider-specific provider.options. The usage object reports exact cost in USD. GPT Image models support SSE streaming previews. Source: X/@OpenRouter thread, 2026-06-24; OpenRouter Blog, 2026-06-23
Current Source Snapshot
The public /api/v1/images/models endpoint returned 38 models across nine provider namespaces when checked on 2026-06-30. The first listed model was openai/gpt-image-2; its record accepts text and image inputs, emits image output, exposes quality/background/count/input-reference/output-compression controls, and reports streaming support. Source: OpenRouter API, 2026-06-30
The docs define model records, endpoint records, and capability descriptors. Endpoint records carry provider slug/tag, supported parameters, allowed passthrough parameters, streaming support, and pricing. Capability descriptor values include enum, range, and boolean forms; if a key is absent, agents should treat the capability as unsupported rather than guessing. Source: OpenRouter image-generation docs, 2026-06-30
The reviewed visual artifact shows the OpenRouter model page for openai/gpt-image-2 with generated output, cost, token count, elapsed time, model modalities, price, context, and release metadata. That makes the page useful for human model selection and cost inspection, not only API discovery. Source: X/@OpenRouter visual artifact, 2026-06-24
Kevin Routing
Use this when building agentic image generation, design asset generation, or product UI that needs dynamic provider selection. It pairs with Frontend and Design Skills for design-artifact workflows, ChatGPT Images 2.0 (gpt-image-2) as a captured model surface, and Browser Testing Skills when implementing app-side generation flows. Source: User bookmark, 2026-06-25
Timeline
- 2026-06-30 | Deep-reviewed Kevin's bookmark artifact and refreshed source authority:
/api/v1/images/modelsreturned 38 image models across nine provider namespaces, with typed model and endpoint capability metadata documented for agent-side validation. Source: OpenRouter API/docs, 2026-06-30 - 2026-06-23 | OpenRouter published the Unified Image API announcement. Source: OpenRouter Blog, https://openrouter.ai/blog/announcements/image-api/
- 2026-06-24 | OpenRouter announced the API on X with an agent-specific capability-discovery thread. Source: X/@OpenRouter, https://x.com/OpenRouter/status/2069799707019215241
- 2026-06-25 | Page created from Kevin's X bookmark sync. Source: raw/x-bookmarks/enriched/2069799707019215241.json