June 2025 Summaries

OpenRouter has introduced a new feature called Presets, enabling users to manage their large language model (LLM) configurations from their dashboard, thus separating these settings from the code. This feature allows users to create, manage, and update model settings, system prompts, and routing rules, offering a streamlined way to handle model selection, provider routing, and generation parameters. Users can easily conduct A/B testing or update prompts dynamically without altering the code, thereby simplifying the complexity of configuration management and allowing developers to focus their codebase on product logic. Presets can be referenced in API calls either directly as a model, with a model override, or through a dedicated field, providing flexibility in their application.

Presets is a new feature from OpenRouter that allows users to manage their Large Language Model (LLM) configurations separately from their code, facilitating the creation, management, and updating of model settings, system prompts, and routing rules through a dashboard interface. This feature simplifies the process of defining model selection, provider routing, and generation parameters, enabling users to conduct A/B testing and update prompts dynamically without altering the code. By focusing the codebase on product logic, Presets offload the complexity of configuration management, offering multiple ways to reference presets in API calls, such as directly using a preset slug or with a model override.

OpenRouter has announced updates to its offerings, including a new Uptime Tracking API that allows users to programmatically monitor model uptime, enhancing transparency and reliability. Additionally, improvements have been made for Bring Your Own Key (BYOK) users, such as the ability to test keys before saving, limit upstream usage per key, and track usage with a specific API response parameter. These enhancements aim to provide users with better control and insight into their API interactions. Links to detailed information about these updates are provided on their platform.

OpenRouter has introduced a new API endpoint for programmatically tracking model uptime, enhancing the ability of developers to monitor system performance efficiently. In addition to this, the platform has implemented several improvements for users utilizing the Bring Your Own Key (BYOK) feature, including the ability to test keys before saving them, limit upstream usage per key, and track usage by indicating "is_byok": true in API responses. These updates aim to provide a more robust and user-friendly experience for developers interacting with OpenRouter's services.

OpenRouter has announced changes to its platform fee structure, eliminating the fixed $0.35 Stripe fee and simplifying the calculation of fees for credit purchases. Under the new structure, non-crypto payments will incur a 5.5% fee with a minimum of $0.80, while crypto payments are subject to a 5.0% flat rate with no minimum fee. Additionally, the current 5% BYOK usage-based fee will eventually be replaced by a fixed monthly subscription, though pricing details are yet to be determined. These adjustments are expected to result in lower overall costs for most users, and no action is required from them as billing updates will occur automatically.

OpenRouter is revising its platform fee structure by eliminating the fixed $0.35 Stripe fee, making fees easier to calculate and generally resulting in lower costs for users. The new fee structure imposes a 5.5% fee with a minimum of $0.80 for non-crypto payments, while crypto payments incur a flat 5.0% fee with no minimum. The existing BYOK fee will shift from a 5% usage-based model to a fixed monthly subscription, with details yet to be determined. Examples show that most users will experience reduced fees, with billing adjustments being automatic, requiring no action from users.

OpenRouter has introduced several new features aimed at enhancing developer experience by making the platform more efficient and user-friendly. Users can now send GIF prompts directly to models from OpenAI, Gemini, Anthropic, and Llama without needing to pre-convert images, improving workflow efficiency. The Omni-Search function allows users to quickly access provider pages by typing a provider's name, providing immediate information on available models, pricing, and uptime. For Anthropic model users, tool-call caching has been implemented, reducing latency and token usage. Additionally, a new is_byok flag in usage responses confirms whether a "bring your own key" setup was used, offering more transparency in API usage.

OpenRouter has introduced new developer features to enhance its functionality, including GIF support across various models, provider pages in Omni-Search, Anthropic tool-call caching, and an "is_byok" flag in usage. Users can now send image or GIF prompts directly to models such as OpenAI, Gemini, Anthropic, and Llama without needing to pre-convert the files. The Omni-Search feature enables users to quickly access provider pages by typing the provider's name, allowing them to view model details, pricing, and uptime. The tool-call caching feature for Anthropic significantly reduces latency and token usage, improving efficiency. Additionally, the introduction of the "is_byok" flag in usage responses confirms whether a bring your own key setup was utilized, providing increased transparency in the usage process.