one year on
OpenAI launches GPT-4.1 family of coding-focused models via API
The new models feature a 1-million-token context window and are priced from $0.10 to $2 per million input tokens, while GPT-4.5 deprecation looms.
OpenAI today released GPT-4.1, a new family of models available exclusively through its API, not ChatGPT. The lineup includes GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano, all boasting a 1-million-token context window capable of processing roughly 750,000 words at once. The company says the models are optimized for coding and instruction following, with improvements in frontend coding, reduced extraneous edits, and consistent tool usage.
Pricing is $2 per million input tokens and $8 per million output tokens for the full GPT-4.1; GPT-4.1 mini costs $0.40/$1.60, and GPT-4.1 nano costs $0.10/$0.40, making it the cheapest model OpenAI has ever released. On SWE-bench Verified, GPT-4.1 scored between 52% and 54.6%, trailing Google’s Gemini 2.5 Pro (63.8%) and Anthropic’s Claude 3.7 Sonnet (62.3%). OpenAI notes the model’s accuracy drops from about 84% at 8,000 tokens to 50% at 1 million tokens.
The launch comes as OpenAI also announced deprecation plans for GPT-4.5, and the naming scheme — 4.1 after 4.5, before o3 — has drawn widespread mockery from developers on social media, with many calling it a new peak of versioning absurdity.
Developers express confusion over the naming scheme, noting GPT-4.1 arrives after GPT-4.5 and before o3, sparking jokes about versioning logic.
One year later — open only if you can handle spoilers
The GPT-4.1 family went on to power a wave of third-party coding agents, though its benchmark gap with competitors persisted. The naming confusion was largely forgotten by late 2025 as OpenAI shifted focus to o-series reasoning models.