A rival implements the Responses API natively, and the interface stops being a moat
DeepSeek's V4-Pro GA includes native support for the OpenAI Responses API, for Codex-style setups. When a competitor implements your interface on purpose, the interface has become plumbing.
The V4-Pro general availability release includes native support for the OpenAI Responses API, aimed at Codex-style setups.
This is the moment an API becomes infrastructure. A competitor does not implement your interface to flatter you — it implements it because the tools its customers already use speak that dialect, and the cost of not speaking it is higher than the cost of adopting a rival's shape.
The precedent is the chat completions endpoint, which every provider now serves whether or not they like it. What is new is that the Responses API is a richer surface — tool calls, state, structured output — so implementing it natively concedes more of the architecture than a completions shim ever did.
For anyone building on models this is unambiguously good and slightly hazardous. Good, because it makes providers swappable at the integration layer. Hazardous, because a shared interface hides real behavioural differences: the same request to two models that both accept it can produce very different tool-calling behaviour, and the interface will not warn you.
It is the same convergence visible in pricing and modality this week — endpoints collapsing into general models, tiers collapsing into parameters. The differentiators are moving off the surface and into what happens behind it.
Digital Applied — AI Model Releases: August 2026 Tracker and Dated Ledger → · How AI Works — Best Open-Weight LLM for Agentic Coding in 2026 →