OpenAI found that GPT-5.6 Sol's surprisingly low ARC-AGI-3 benchmark score was caused by the test harness dropping reasoning history and aggressively truncating context. By switching to its Responses API and enabling retained reasoning alongside compaction, the model tripled its score while slashing output tokens by 6x. It is a stark reminder that benchmarks evaluate harness architecture just as much as raw intelligence.
OpenAI just refreshed its official prompt engineering guide, laying out best practices for message roles and prompt structuring in the Responses API. The guide urges developers to version prompts directly inside application code, noting that hosted prompt objects will be phased out by November 2026.
OpenAI just rolled out WebSocket mode for its Responses API, tackling the cumulative network and validation overhead that drags down fast agent loops. By keeping a persistent connection and caching previous response states in memory, end-to-end latency drops by up to 40% with a 45% faster time to first token. Coding tools like Cursor and Vercel are already clocking major speedups.
OpenAI just upgraded its Responses API with a native shell tool, hosted container workspaces, and built-in context compaction for GPT-5.2 and later models. Agents can now run Unix commands like curl or grep concurrently, stage files, and query SQLite databases in sandboxes. It basically saves developers from having to build their own execution harness for long-running workflows.
OpenAI just made its specialized coding model, GPT-5.3-Codex, available to all developers through the Responses API. The model combines frontier coding performance with deep professional knowledge in a single package. It basically gives builders immediate programmatic access to integrate its top-tier code generation directly into their own apps.
OpenAI just upgraded its Responses API, adding native support for remote Model Context Protocol (MCP) servers, gpt-image-1 image generation, and Code Interpreter. Reasoning models like o3 and o4-mini can now call tools directly inside their chain-of-thought while preserving context tokens. OpenAI also officially joined the MCP steering committee to back the open standard.
OpenAI just rolled out direct PDF inputs for its Responses and Chat Completions APIs. Developers can now pass raw PDF documents straight to models like o1, gpt-4o, and gpt-4o-mini without third-party parsing tools. Why it matters: document workflows just got a lot simpler.
Microsoft just rolled out the Responses API and Computer-Using Agent (CUA) inside Azure AI Foundry. CUA lets AI autonomously navigate software interfaces, click buttons, and fill forms from natural language prompts, while the new API bundles tool calls into a single request. It turns AI from an advisor into an operator, though human oversight is still urged for sensitive workflows.
OpenAI just launched a dedicated suite of tools to help developers build AI agents, anchored by the brand-new Responses API alongside built-in web search, file search, and computer use. It also released an Agents SDK and tracing tools to orchestrate complex multi-step workflows. Why it matters: OpenAI is streamlining agentic apps into a single endpoint and plans to sunset the Assistants API by mid-2026.
