Added Llama 4 to routable models
Meta's Llama 4 is now available through /v1/auto/route for text, with both instruction-tuned and reasoning variants.
Changelog
Every model, capability, and platform change from the GreatRouter team — published as it ships.
New cost optimization reports
The dashboard now breaks down spend by model, provider, and task type with daily, weekly, and monthly views. Export to CSV is one click.
Enterprise tier launch
Volume discounts, committed-use pricing, SSO, audit logs, and a dedicated engineering contact. Email contact@greatrouterai.com for details.
Native speech-to-text routing
ASR is now a first-class routed task. The classifier detects audio inputs and routes to the best transcription model for the language and audio profile.
Health-aware failover improvements
Per-model error rates now flow into the scoring engine in real time. The router avoids degraded models automatically and surfaces the cause in your activity log.
GPT-5 added to routable models
OpenAI's GPT-5 is available via /v1/auto/route and /v1/chat/completions. Pin a specific model id or let the router pick based on task and budget.
Per-organization routing preferences
Set a default optimization mode, exclude specific providers, and prefer specific models — per organization, with API and dashboard management.
Budget caps per request
The budget_dollars parameter now hard-stops the router from considering models above the cap. Works across text, image, video, and audio requests.
Regional data residency in EU
New EU data residency option on enterprise plans. Billing metadata and request logs are stored in Frankfurt; provider calls remain closest-region.
Flux 2 Pro integration
Black Forest Labs' flagship image model is now in the registry. Supports photorealistic, stylized, and brand-LoRA workflows through the same endpoint.
Auto-recharge thresholds
Configure auto-recharge with configurable thresholds and top-up amounts. Wallet events are now visible in the activity log.
Suggest endpoint for human-in-the-loop
POST /v1/auto/suggest returns the top-ranked models for a task without executing the request. Useful for high-stakes flows that need approval before inference.
Claude Sonnet 4 support
Anthropic's Sonnet 4 is in the registry with both text and vision modalities. Available via auto-route and the OpenAI-compatible endpoint.
GreatStudios integration
GreatStudios now uses GreatRouter as its default inference layer. Every creative tool — image, video, music, chat — routes through the same intelligent endpoint.
Improved prompt enhancement for short inputs
Image, video, music, and code prompts under 120 characters are auto-expanded using a small reasoning model before being sent to the upstream provider.
Subscribe to updates
Each entry here also ships to the newsletter. Subscribe from the blog or follow the public changelog RSS feed (coming soon).