The Cloudflare Blog: Partnering with OpenAI to bring their new open models onto Cloudflare Workers AI

Aug 6, 2025

—

Source URL: https://blog.cloudflare.com/openai-gpt-oss-on-workers-ai/
Source: The Cloudflare Blog
Title: Partnering with OpenAI to bring their new open models onto Cloudflare Workers AI

Feedly Summary: OpenAI’s newest open-source models are now available on Cloudflare Workers AI on Day 0, with support for Responses API, Code Interpreter and Web Search (coming soon).

AI Summary and Description: Yes

**Short Summary with Insight:**
This announcement highlights Cloudflare’s collaboration with OpenAI to integrate open-weight AI models into their Workers AI platform. It represents a significant advancement in the availability of open-source AI capabilities for developers, enhancing transparency, flexibility, and security in deploying AI applications. The new models are particularly aimed at organizations that prioritize data security and privacy, making them highly relevant for professionals focused on compliance and secure infrastructure.

**Detailed Description:**

– **Partnership Announcement:** Cloudflare is celebrating its partnership with OpenAI as a Day 0 launch partner, offering access to OpenAI’s first open-weight models on the Cloudflare Workers AI platform.
– **Model Sizes and Specifications:** The new models come in two sizes: a 120 billion parameter model and a 20 billion parameter model, both utilizing a Mixture-of-Experts (MoE) architecture. This architecture enables faster and more memory-efficient performance compared to traditional dense models.
– **Quantization:** The models run at FP4 quantization, significantly reducing GPU memory usage.
– **Capabilities:** Despite being text-only, these models possess advanced reasoning capabilities, as well as features for code interpretation and web search.
– **Response APIs:** The integration includes support for the new Responses API format, promoting an effective means of interaction with the models.
– Various REST API endpoints are available for executing model interactions, enhancing usability for developers.
– **Code Interpreter Utilization:** The announcement emphasizes a stateful Code Interpreter integrated with Cloudflare’s Sandbox environment, which supports secure and dynamic code execution.
– This feature is particularly advantageous for tasks requiring logical responses, such as coding or mathematical computations, addressing common difficulties faced by large language models (LLMs).
– **Infrastructure Advantages:** Cloudflare’s infrastructure allows for the running of full applications built on their developer platform, providing an ecosystem that enhances the utility of the AI models.
– **Call to Action:** Developers are encouraged to explore the new models’ capabilities through Cloudflare’s Developer Docs, reflecting an innovative approach to disseminating advanced AI tools.

In summary, this integration not only provides performance benefits through advanced model architecture but also prioritizes security and flexibility, which are critical considerations for AI deployments in business environments. The insights gained from this development are essential for security and compliance professionals looking to harness AI effectively while ensuring data protection and operational integrity.

1 2 4 a access Act actions advanced advanced AI advanced reasoning Advanced Reasoning Capabilities advancement age AI AI applications AI capabilities ai model AI models AI tool AI tools and anti API APIs app Application applications Arch architecture art as at ated availability being benefits Bi Box built business by C C code capabilities CI Cloud Cloudflare Cloudflare Workers co code code execution Code Interpreter coding Col collaboration compliance compliance professionals computation critical D data Data Protection data security day de deployment deployments developer developer platform developers development e ecosystem effective efficient end endpoint endpoints environment ERP execution exp expert Experts Experts (MoE) face fast faster feature features first flexibility focused for full g geo GPT GPU H high Highlight HR http HTTPS in infrastructure innovative approach insights integration integrity inter interaction interactions interpret interpreter io Iron IRS ite k l Labor language language model language models large large language model large language models Large Language Models (LLMs) led Li llm llms lm logic low M making man math mathematical mean memory memory usage Mixture mixture-of-experts Mode model model architecture models MoE N new no o oE of off on only ons open open models open-source open-source models openai operation operational operational integrity OPM organization organizations ory oS oss parameter partners partnership per performance performance benefits platform point pre privacy pro professionals protection ps Q quantization R rag rate RCE re reasoning reasoning capabilities red response responses Ro RoT s sandbox sandbox environment search sec secure secure infrastructure security security and compliance short side Sig size sizes source source model source models specific SSE state support system T Task tasks ted text the to tool tools TP transparency two UI up US usability usage use utilization V Vantage web web search weight weight models Well Wi workers x z