Recently, Microsoft announced the release of GPT-5 in Azure AI Foundry.
Like any new technology, GPT-5 has generated a great deal of feedback, mostly because expectations were high for the latest iteration of the model that captured both the imagination and a significant user base for this generation of AI offerings.
In Azure AI Foundry, the GPT-5 models are available via API and orchestrated by the model router. The GPT-5 series spans complementary strengths:
- GPT-5, a full reasoning model provides deep, richer reasoning for analytics and complex tasks, like code generation, with a 272k token context.
- GPT-5 mini powers real-time experiences for apps and agents that require reasoning, tool calling to solve customer problems.
- GPT-5 nano is a new class of reasoning model which focuses on ultra-low-latency and speed with rich Q&A capabilities.
- GPT-5 chat enables natural, multimodal, multi-turn conversations that remain context-aware throughout agentic workflows, with 128k token context.
Together, the suite delivers a seamless continuum from rigorous agentic coding tasks, to relatively simple Q&A—all delivered with the same Azure AI Foundry endpoint using model router in Foundry Models.
—GPT-5 in Azure AI Foundry: The future of AI apps and agents starts here, Microsoft Azure blog
Now that the initial feedback wave has receded a bit, we thought it might be helpful to offer some of our favorite—and most applicable—features for startup founders to consider as they’re exploring the benefits of GPT-5 on Azure.
Our internal AI experts at Microsoft for Startups were eager to share their takes GPT-5. And we even got one of our startups to chime in.
Here’s what they had to say.
GPT-5 on Azure: Smarter approach to model routing
Insight from Amit Svarzenberg, Chief Technology Officer at Microsoft for Startups
One of the most compelling advantages of GPT-5 on Azure is its intelligent model routing. Unlike OpenAI’s native setup, Azure’s Model Router supports cross-family routing across GPT-5, GPT-4.1, o-series, and reasoning models. This means developers can deploy a single endpoint that automatically selects the best model for each request—balancing cost, speed, and quality. For startups, this translates into smarter model selection, simplified integration, and built-in governance and monitoring—all while staying up to date.
Get more of Amit’s perspective on GPT-5 on Azure.
GPT-5 on Azure: Purpose-built for agentic AI workflows
GPT-5 on Azure is designed with agentic AI in mind. It allows for high-level workflows that rely on reasoning and tool use, making it ideal for startups building agent-based systems. With support from the Azure Marketplace, startups can deploy agents that are both cost-effective and powerful. GPT-5 enhances these workflows by reducing hallucinations and deception rates, both aspects that are critical for production environments.
Safety is also a standout feature. Microsoft’s AI Red Team found GPT-5 to be qualitatively safer than OpenAI’s o3 model, especially in frontier and content safety domains. It resists jailbreaks and refuses to generate offensive or weaponizable code. These safeguards make GPT-5 a reliable choice for startups operating in sensitive or regulated domains.
Beyond the agentic workflows, GPT-5 introduces robust safety improvements. It has stronger content moderation, and in psychosocial contexts, GPT-5 shows improved detection of emotional distress, which is an area where both Microsoft and OpenAI have seen promising results.
Additional features include guardrails against bio-weaponization, reduced sycophancy, and the ability to process much longer documents such as books, meeting transcripts, and datasets. These enhancements make GPT-5 a versatile and secure solution for startups with complex data needs.
GPT-5 on Azure: Real-world validation from a startup perspective
Feedback from DryMerge, a workflow automation startup
Microsoft folks being excited about Microsoft offerings is one thing. But what about the startups using this technology in their day-to-day operations? We turned to DryMerge, a startup in the Microsoft for Startups program. They’ve been able to use GPT-5 to drastically improve their CRM automation—without even changing their prompts. Here’s what they highlighted that may be useful as you’re building your startup:
- Streamlined searching: GPT-5 explores every angle—alternate spellings, partial matches, and indirect connections—like someone who genuinely wants to find the answer.
- Handling ambiguity: It makes sense of messy, real-world data. Custom fields with varying meanings across contexts? GPT-5 figures it out.
- Complex instruction execution: DryMerge tested GPT-5 with more than 200 business rules, including exceptions and hierarchical logic. The model followed them all without explicit instruction.
- Context awareness: GPT-5 adapted to custom business logic that would typically require hours of manual configuration.
For more on DryMerge’s experience with GPT-5, see this post from Sam Brashears, Founder and Chief Technology Officer of DryMerge.
Want to get started?
With Azure AI Foundry’s first-class reliability, real-time evaluations, built-in observability, and secure deployment options, you can confidently move from pilot to production—all aided while unique tools like Model Router optimizes quality, latency, and cost across workloads.
—GPT-5 in Azure AI Foundry: The future of AI apps and agents starts here, Microsoft Azure blog
If you’re a startup interested in implementing GPT-5, check out this guide on OpenAI prompt optimization for practical tips on getting the most out of your deployment.
