GPT-4.1

weights

closed

data

closed

code

closed

license

Proprietary(API-only)

GPT-4.1 is a proprietary OpenAI API model with no weights, training data or code released. OpenAI's own documentation puts its context window at 1,047,576 tokens.

• https://developers.openai.com/api/docs/models/gpt-4.1 recorded 2026-09-29

  OpenAI model page for GPT-4.1 listing a "1,047,576 context window" and the snapshot "gpt-4.1-2025-04-14"; searched the body for open-weight, weights download, source code and license terms: none, the model is served only through the API.

• https://techcrunch.com/2025/04/14/openais-new-gpt-4-1-models-focus-on-coding/ recorded 2026-09-29

  TechCrunch report that the GPT-4.1 models "are exclusively available through the company" API; searched the body for open weights, open source and license: none.

GPT-4.1 launched as an API-first developer and coding model, initially outside ChatGPT's default lineup, and was adopted as a cheaper replacement for GPT-4o across developer tooling. OpenAI discloses no per-model usage count, so this reading comes from its broad developer-platform distribution as a major API model rather than a measured figure.

• https://techcrunch.com/2025/04/14/openais-new-gpt-4-1-models-focus-on-coding/ recorded 2026-08-13

  GPT-4.1 positioned as the API coding family replacing GPT-4o; no usage or user figure published anywhere on the page

GPT-4.1 scored 54.6% on SWE-bench Verified at its April 2025 release, 21 points above GPT-4o at 33.2%. It is a non-reasoning model, and OpenAI's newer GPT-5 series has since replaced it as the main production line.

• https://www.helicone.ai/blog/gpt-4.1-full-developer-guide recorded 2026-09-29

  Helicone guide says GPT-4.1 "scored 54.6%, far outperforming GPT-4o (33.2%) and GPT-4.5 (28%)" on SWE-bench Verified.

• https://www.infoq.com/news/2025/05/openai-gpt-4-1/ recorded 2026-09-29

  InfoQ write-up: "54.6% accuracy. This is a 21-point increase over" GPT-4o (33.2%).