An analyst describes the latest launch as a “defensive move” aimed at staying competitive with AWS, Google, and Microsoft in the AI development space.
OpenAI’s Tuesday release of OpenA1 o1, its reasoning model via API, along with three new developer tools, is a “smart move” to stay relevant in the corporate and enterprise space as more customers prioritize agentic applications over basic chat-based ones, according to an industry analyst.
Jason Andersen, VP and principal analyst at Moor Insights & Strategies, called it “a somewhat defensive move,” noting that AWS, Google, and Microsoft have been rolling out API-based multi-model development frameworks like Bedrock and AI Foundry.
He noted that some of these frameworks “include OpenAI as a potential model option but also introduce features like model routing, meaning there’s no assurance the framework provider will consistently rely on OpenAI.“
As part of its 12 Days of Shipmas campaign, OpenAI announced that OpenA1 o1 in API is now production-ready. The model features vision capabilities that enable it to “reason over images,” opening up applications in fields like science, manufacturing, and coding, where visual inputs are critical. Additional upgrades include function calling for connecting o1 to external data and APIs, as well as a new parameter allowing developers to “control how long the model thinks before answering.” It is currently available to Tier 5 users in the API.
The toolset also includes upgrades to the Realtime API, which OpenAI stated “enables developers to build low-latency, multi-modal conversational experiences.” It supports text and audio inputs and outputs, along with function-calling capabilities.
A standout feature is the integration with WebRTC, an open standard that simplifies the development and scaling of real-time voice products across platforms. “Our WebRTC integration is designed to deliver seamless and responsive interactions, even under variable network conditions,” the company stated.
Additionally, OpenAI introduced Preference Fine Tuning, a new customization technique that allows models to be tailored more easily to user and developer preferences, as well as beta versions of new SDKs for Go and Java.
Andersen called the new SDKs “quite interesting,” noting that Go and Java are widely used by developers working with public cloud platforms. He added that these SDKs and APIs offer a significant alternative to relying solely on pre-built models from cloud providers.
In contrast, Forrester senior analyst Andrew Cornwall remarked, “When OpenAI says ‘new tools for developers,’ they mean tools for AI developers. The broader audience will only benefit once AI developers start leveraging these tools, which could take time, especially since the o1-2024-12-17 APIs are currently limited to Tier 5, OpenAI’s highest-usage users.“
He noted that, based on the announcements, “the most notable addition is vision support, a feature not included in o1-preview.” Additionally, the latest o1 model is expected to provide more accurate responses and experience fewer refusals compared to the preview version.
Regarding math and coding capabilities, Cornwall said, “These have improved since the preview, in some cases significantly, although o1’s reported score won’t surpass Amazon Q as the leader of the SWE-bench Verified leaderboard.” OpenAI has also enhanced function calling and structured output, which allows for calling into a program from an AI query and returning results in developer-friendly JSON instead of natural language.
He also highlighted that “[developers now have] a new parameter to experiment with, reasoning effort, which allows them to balance accuracy and speed.” He added, “This will work well with WebRTC voice support—while a brief pause might be fine when typing, it’s not ideal during a conversation.”
He also mentioned, “More developers will have the opportunity to work with OpenAI’s most powerful model, as SDKs for Java and Go transition from alpha to beta. Businesses are eager to adopt AI but have been waiting for the stability that a beta version provides.“
“In summary,” Cornwall stated, “this isn’t a revolutionary change—many of these capabilities have already been available in different models. However, with this release, a single model improves in various modes and for a wider range of problems.“
He also mentioned, “More developers will have the opportunity to work with OpenAI’s most powerful model, as SDKs for Java and Go transition from alpha to beta. Businesses are eager to adopt AI but have been waiting for the stability that a beta version provides.“
“In summary,” Cornwall stated, “this isn’t a revolutionary change—many of these capabilities have already been available in different models. However, with this release, a single model improves in various modes and for a wider range of problems.”
Â
Â
Â