OpenAI announced the release of a new family of artificial intelligence models designed to optimize coding capabilities. This launch is part of OpenAI’s strategy to compete with other tech companies such as Google and Anthropic. The models are accessible to developers through OpenAI’s application programming interface (API).
The lineup includes three model sizes: GPT 4.1, GPT 4.1 Mini, and GPT 4.1 Nano. OpenAI’s Chief Product Officer, Kevin Weil, stated during a livestream that these new models outperform OpenAI’s most popular model, GPT-4o, and surpass the capabilities of its largest model, GPT-4.5, in certain aspects.
The GPT-4.1 model achieved a 55 percent score on the SWE-Bench, a prominent benchmark used to evaluate coding model capabilities. This score exceeds that of OpenAI’s previous models. Weil commented that the new models excel in coding, complex instruction following, and are highly effective for developing AI agents.
Recent advancements in AI models have greatly enhanced their ability to write and edit code, allowing for more automated software prototyping and improving the functionality of AI agents. Competitors like Anthropic and Google have also introduced models with strong coding capabilities in recent months.
Speculation about the release of GPT-4.1 had been circulating, with reports that OpenAI tested the model under the pseudonym Alpha Quasar. Users who tried this “stealth” model shared positive feedback regarding its coding proficiency, with one Reddit user noting that it resolved issues they encountered with code generated by other models.
Michelle Pokrass, part of OpenAI’s post-training team, explained during a livestream that developers prioritize coding and that OpenAI has been enhancing its model’s ability to generate functional code. Efforts have focused on improving the models’ adherence to formats, repository navigation, unit testing, and code compilation.
OpenAI has capitalized on the interest generated by ChatGPT, a chatbot launched in late 2022, to build a business around more advanced chatbots and AI models. In a TED interview, CEO Sam Altman revealed that OpenAI had 500 million weekly active users, with rapid growth in usage.
The company offers a variety of models with different capabilities and pricing structures. Its most powerful model, GPT-4.5, was introduced as a “research preview” earlier this year. Additionally, models such as o1 and o3 are designed for simulated reasoning, though they have longer response times and higher user costs.
The success of ChatGPT has led to a proliferation of similar products, motivating competitors to increase their research investments to catch up with OpenAI. A recent report from Stanford University highlighted that models from companies like Google, DeepSeek, Anthropic, Meta, and Mistral are now approaching the capabilities of OpenAI’s offerings.