Zenith and Prism (and the upcoming Ember) now run on Gemma 4 26B (256K context), after our upstream provider retired Mistral Nemo. Requests also fail over automatically to Llama 3.3 70B if a model's upstream is unavailable
Model cards now show accurate context windows
Platform launch
New
July 2026
OpenAI-compatible API with 8 specialized writing models (Lyra, Zenith, Echo, Verse, Prism, Sage, Nova, Atlas)
Streaming chat completions with exact token metering
In-browser playground with model comparison mode and code export
API key management with reveal-once secrets and monthly spend caps
Per-key rate limiting with x-ratelimit response headers
Usage analytics dashboard with per-model and per-key breakdowns, CSV export
Pay-per-use credit billing with free starter credits on account activation
Self-serve card payments for credit packs via Polar