Changelog

​

March 2026

​

v1.4.0 — Model Comparison API

• New: GET /v1/models/compare endpoint for side-by-side model evaluation
• New: Quality scores (overall, code, math, reasoning) in model responses
• New: Performance metrics (latency p50/p99, availability) per model
• Improved: Model listing now includes pricing breakdown (official vs OpenModex rates)

​

v1.3.0 — BYOK Support

• New: Bring Your Own Key — use your own provider API keys
• New: AWS KMS envelope encryption for provider key storage
• New: Provider key management in the Dashboard
• Improved: Routing engine handles mixed BYOK + managed configurations

​

February 2026

​

v1.2.0 — React Hooks

• New: @openmodex/react package with useChat, useCompletion, useModels hooks
• New: Built-in streaming and abort support in React hooks
• Improved: TypeScript SDK now exports all types

​

v1.1.0 — Prompt Caching

• New: Server-side prompt caching with configurable TTL (60s–24h)
• New: cache_hit field in response metadata
• Improved: Cached responses billed at 50% input rate, zero output cost

​

January 2026

​

v1.0.0 — General Availability

• Launch: OpenModex API Gateway — OpenAI-compatible API for 100+ models
• Launch: Smart routing (cost, latency, quality strategies)
• Launch: Automatic fallbacks with circuit breaker
• Launch: Official SDKs for Node.js, Python, and Go
• Launch: CLI tool (@openmodex/cli)
• Launch: Dashboard with usage analytics and billing
• Providers: OpenAI, Anthropic, Google Gemini, DeepSeek, Mistral, Qwen