Multi-Region Routing, Content Filters & More
Route requests to regional providers, protect your apps with built-in content moderation, enforce API key rate limits, and explore new models.
Multi-Region Provider Routing
LLM Gateway now supports regional routing — send requests to geographically specific provider endpoints for lower latency and data residency compliance.
- Route to specific regions per provider (e.g.
us-east,eu-west) - Region-specific pricing automatically applied
- Regional mapping statistics tracked separately
- Fallback to root mappings when regional endpoints are unavailable
Content Filters & Moderation
Built-in content moderation gives you control over what passes through your gateway:
- Configurable moderation modes — choose between blocking, logging, or passthrough
- Content filter routing — automatically route requests based on content classification
- Provider exclusion — exclude providers from content-filtered requests
- Moderation payloads stored for audit and review
- Separate image moderation pipeline for multimodal requests
API Key Period Limits
Set time-based rate limits on individual API keys — cap usage by hour, day, or month to control spend and prevent runaway costs.
Provider RPM Caps
Configure requests-per-minute caps at the provider level. When a provider hits its RPM limit, requests automatically fall back to the next available provider.
New Models
MiniMax Reasoning
MiniMax models now support split reasoning — separating thinking tokens from response tokens for better transparency and cost tracking.
Token Cost Calculator
New Token Cost Calculator tool — estimate costs across models and providers before committing to a workflow.
UI & Platform Improvements
- Redesigned provider page — Model cards on provider pages rebuilt with clearer pricing, status badges, and capability tags
- Diff syntax highlighting — Code blocks now render diffs with proper
+/-line coloring - Model CTA — Contextual call-to-action on model pages to guide new users
- Admin log filters — Improved filtering in the admin dashboard activity logs
- Credit top-up cap — Maximum cap on credit top-ups for safer billing
- Auto top-up expiry — Auto top-up now disables after 7 days of inactivity
Fixes & Reliability
- OpenAI streaming keepalive events no longer cause parsing errors
- OpenAI response done events handled correctly in streaming
- Streamed provider errors now surface proper error messages
- ByteDance sensitive content blocks properly classified
- Moonshot thinking disabled for tool calls to prevent errors
- Regional gateway routing simplified for better reliability
- E2E tool call test failures resolved