LoginSign Up
← Back to all sparks
O

OpenRouter

AI-ASSISTANTS
Velocity8.8

Unified API marketplace for 200+ LLMs from OpenAI, Anthropic, Google, Mistral, Meta and more.

openrouter.ai ↗

OpenRouter launches US in-region data routing, completing its compliance story for regulated industries.

model-routingdata-residencyagentic-toolingcompliancemulti-model
◆Current state
OpenRouter is expanding rapidly from a model-routing API to a full AI application infrastructure layer. The past two weeks brought US in-region routing (joining EU), Fusion compound model routing, config-as-code Presets, Zero Data Retention documentation, and a shell server tool with Files API. The product changelog doubles as technical documentation — entries are detailed guides, not just feature announcements.
◆Where it's heading
OpenRouter is building the infrastructure that makes AI models interchangeable and compliance-safe for enterprise applications: in-region routing for data geography, presets for configuration management, Fusion for quality, and shell execution for agentic workloads. The consistent direction is: make any model work anywhere, safely, with minimal integration code.
◆Prediction
The shell server tool will move from beta to GA and gain persistent sessions. Presets will likely get versioning and rollback. Compliance features (SOC 2, HIPAA attestations) are the natural next step following in-region routing, targeting the regulated industries these features unlock.

◆Recent moves

  1. 5d ago

    Zero Data Retention (ZDR): What It Means for AI APIs

    Zero Data Retention is now formally documented with clear boundary conditions — what ZDR covers, what it doesn't, and how to enforce it at the account, guardrail, or request level. Signals OpenRouter is investing in the compliance narrative required to win enterprise customers evaluating AI vendor data handling.

  2. 6d ago

    How to Use OpenRouter Presets: Config-as-Code Guide

    Presets are named, versioned configurations that encapsulate model selection, system prompts, provider routing, and sampling parameters — referenced as @preset/your-slug and updated from the dashboard without redeploying applications. This decouples model configuration from code, a practical improvement for teams managing multiple AI applications.

  3. 6d ago

    OpenRouter Fusion: How It Works and When to Use It

    OpenRouter Fusion routes a single prompt through multiple models in parallel, uses a judge to map agreement and disagreement, and lets the calling model write the final answer. Quality improves at the cost of latency and tokens — appropriate for high-stakes decisions, not real-time UX. Documents trade-offs clearly.

  4. 7d ago

    Nano Banana API: Edit Images with Gemini in Code

    A developer guide for Gemini-powered image editing via OpenRouter's Nano Banana API — one request, source image in, edited image out, model swappable by changing a single field. Runnable Python and TypeScript examples. Documents a real capability with enough detail to be actionable.

  5. 7d ago

    Seedance 2.5 Review: What It's Best At and When to Use It

    A model review comparing Seedance 2.5 against Seedance 2.0, Wan 3.0, and Veo 3.1 across resolution, length, and cost. Informational content for model selection — no product capability change.

  6. 7d ago

    In-Region Routing: Keep your data in the US or EU

    ⚡ SPARK

    US in-region routing joins the EU equivalent: requests to us.openrouter.ai are decrypted and served exclusively by US-based providers, with no cross-region transit, across models from OpenAI, Anthropic, Google, and others. Completes the geographic data residency picture that enterprise and regulated-industry customers require.