OpenAI Responses API gets a major upgrade instead of Assistants API
OpenAI has rolled out a significant upgrade to its Responses API, introducing a suite of powerful tools and enterprise-grade features that transform how developers build agentic applications. Announced on May 21, 2025, this release builds upon the initial Responses API launched in March 2025, which replaced the Assistants API and has already processed trillions of tokens across models like GPT-4o and the o-series reasoning engines.
How It Differs from the Traditional ChatGPT (Chat Completions) API
- Agent-First Primitive vs. Text-Only Completions: Unlike the Chat Completions API, which returns plain text based on prompts, the Responses API is designed as a core primitive for building “agentic” experiences—allowing models to plan and execute multi-step tasks by calling external tools directly within their chain-of-thought.
- Built-In Tool Orchestration: While Chat Completions offers function-calling capability, Responses unifies tool invocation—such as image generation or code execution—into a single, streamlined API call, reducing boilerplate and improving developer productivity.
- Preserved Reasoning State: Models like o3 and o4-mini maintain reasoning tokens across calls and tool invocations, yielding richer contextual understanding and lower latency compared to stateless completions.
- Enterprise-Grade Reliability: Features such as background mode for asynchronous tasks, reasoning summaries for auditability, and encrypted reasoning items for Zero Data Retention customers deliver stronger SLAs and privacy controls than the standard Chat Completions endpoint.
New Capabilities
- Remote MCP Server Support: Connect any Model Context Protocol server—Shopify, Stripe, Twilio, and more—to extend model context with third-party data sources via just a few lines of code .
- Native Image Generation: Access the
gpt-image-1model as a tool within Responses, enabling streamed previews and multi-turn edits without separate API calls . - Integrated Code Interpreter: Perform data analysis, complex computations, and image manipulations directly within the agentic flow, boosting performance on industry benchmarks .
- Enhanced File Search: Query across multiple vector stores with attribute filters to pull relevant document snippets into context, simplifying knowledge-base integrations.
- Enterprise Features: Background mode to manage long-running reasoning tasks, automatic reasoning summaries for debugging, and encrypted reasoning items for compliant deployments .
Pricing and Availability
All new tools and features are available immediately in the Responses API for GPT-4o, GPT-4.1, and the o-series models (o1, o3, o3-mini, o4-mini); image generation is supported on o3 only. Pricing remains consistent with existing tool rates:
- Image Generation: $5.00 per 1 M text input tokens, $10.00 per 1 M image input tokens, $40.00 per 1 M image output tokens (75% off cached inputs)
- Code Interpreter: $0.03 per container execution
- File Search: $0.10 per GB of vector storage per day(first GB free); $2.50 per 1 K tool calls
- Remote MCP Servers: No extra fee—standard output token billing applies.
No separate Responses API fee—tokens are billed at the input/output rates of the selected model (e.g., GPT-4.1 at its published per-token rates).
Developers and enterprises can begin integrating these capabilities today via the client.responses.create endpoint. With these enhancements, OpenAI aims to empower more intelligent, reliable, and secure AI-driven applications across industries.The Responses API is live for all developers today, and the legacy Assistants API will be fully deprecated by mid-2026. Existing Assistants integrations can be migrated with minimal code changes, thanks to compatible request and response schemas .
Getting Started
CometAPI provides a unified REST interface that aggregates hundreds of AI models—including ChatGPT family—under a consistent endpoint, with built-in API-key management, usage quotas, and billing dashboards. Instead of juggling multiple vendor URLs and credentials.
Developers can access latest chatgpt API GPT-4.1 API through CometAPI. To begin, explore the model’s capabilities in the Playground and consult the API guide for detailed instructions. Before accessing, please make sure you have logged in to CometAPI and obtained the API key.
