OpenAI releases GPT-4.1 and the model-naming chapter starts to look ridiculous
A three-model release at lower prices and longer contexts, with naming so confusing the launch post had to include a 'which model should I use' diagram.
On 14 April 2025, OpenAI released GPT-4.1, GPT-4.1 mini and GPT-4.1 nano, a three-tier API-only update to the GPT-4-class line-up. The headline features were a one-million-token context window, lower prices across all three sizes, and improved coding-benchmark numbers, including a step up to fifty-five per cent on SWE-bench Verified for the full GPT-4.1.
What the release also produced, almost immediately, was naming bewilderment. As of mid-April 2025, OpenAI's model menu included GPT-3.5 (legacy), GPT-4 (legacy), GPT-4 Turbo, GPT-4o, GPT-4o-mini, GPT-4.1, GPT-4.1 mini, GPT-4.1 nano, o1-preview, o1, o1-mini, o1-pro and o3 across various tiers. Sam Altman, in a same-day post, acknowledged the issue and announced that GPT-5, when it shipped, would unify the line-up and retire most of these names. The unification took until August.
What GPT-4.1 actually offered
The substantive interest in the release was not the capability frontier. It was the price-per-token and the long-context window. As Wired and Ars Technica wrote in coverage that week, the GPT-4.1 mini at forty cents per million input tokens was, on most benchmarks, comparable to GPT-4o at two-fifty per million. The price-performance line had moved meaningfully in the eleven months since GPT-4o.
The naming was the controversy. The pricing was the news.
The retrospective view of GPT-4.1, four months later when GPT-5 shipped, is that it was a price-cut and quota-bump release misframed as a generational change by its naming. The four hundred-thousand-token to one-million-token context jump was the most operationally consequential improvement, and one that landed quietly at a moment when the field's attention was on reasoning.
Discussion