OpenAI has released GPT-4.1 to API customers, marking the first numbered iteration of GPT-4 since the model family's launch in March 2023 [source]. The company describes the update as improving reasoning, coding, and instruction-following capabilities, though no benchmark comparisons or technical details accompany the announcement.

The release arrives without prior notice to developers. OpenAI's changelog confirms GPT-4.1 is available immediately via the API under the model identifier `gpt-4-1`, with no migration path specified for applications currently calling `gpt-4` or snapshot identifiers like `gpt-4-0613`. The company has not stated whether existing model names will redirect to GPT-4.1 or remain pinned to earlier weights.

No information is provided on training data cutoff, context window size, or whether the model shares architecture with GPT-4 Turbo variants. The announcement does not address whether GPT-4.1 will replace GPT-4 in ChatGPT or remain API-exclusive.

OpenAI's post includes sample outputs demonstrating multi-step reasoning and code generation, but does not publish evaluation results, third-party audits, or comparisons against GPT-4, GPT-4 Turbo, or o1-preview. The company states the model "builds on the foundation of GPT-4" but does not clarify whether this constitutes a retrain, fine-tune, or architectural revision.

Developers report no advance communication via email or the developer forum prior to the changelog appearing. The release follows OpenAI's pattern of unannounced model updates, previously observed with GPT-3.5 Turbo revisions and GPT-4 Turbo snapshot releases. API customers testing production workloads against fixed model identifiers remain unaffected unless they explicitly migrate to the new name.