Aptible LLM Gateway v2.0 is now available to all LLM Gateway customers. This new version is a full rewrite, focused on core improvements:

• Reduced latency: v2.0 adds a baseline response overhead of 2ms on top of LLM provider latency, 5x faster than the 10ms baseline of v1.0. End-to-end latency for applications should see similar improvement

• Automated provider failover (👉 docs)

• Client-side configurable model fallbacks (👉 docs)

• Strict compatibility mode for OpenAI and Anthropic APIs (👉 docs)

• Dedicated endpoints for supported agent harnesses and coding assistants, to address common application-specific errors (👉 docs)

• Always up-to-date model catalog: previously some models listed on our docs or in the UI would not actually be supported by the Gateway, and vice versa. Now, the Supported Models in docs and UI are always 100% in sync with what’s actually supported. This also means no gaps in Model Access Policies that you define.

How to upgrade: See our Upgrade Guide from v1.0 to v2.0.