OpenAI GPT-4 Turbo Preview

gpt-4-turbo-preview was the name developers used for GPT-4 Turbo before the finished model shipped. The line began at OpenAI's first DevDay in November 2023 with the snapshot gpt-4-1106-preview; on 25 January 2024 the alias was pointed at gpt-4-0125-preview, which stayed its default until the end. The full release, gpt-4-turbo, followed on 9 April 2024, so for roughly five months the preview alias was the only way to reach a 128,000-token GPT-4. The specification is the one GPT-4 Turbo shipped with: a 128,000-token context window against the 8,192 of the original GPT-4, at most 4,096 tokens per response, a knowledge cutoff of 1 December 2023, and text in, text out. Pricing never moved from $10 per million input tokens and $30 per million output — sixty-six times what GPT-4o mini charges for input, on OpenAI's own comparison table. One oddity survives in the documentation: the model's snapshot list still includes gpt-4-1106-vision-preview, a vision build filed under an alias whose own card declares text as the only input modality. It is a reminder that during this period the vision, text and function-calling branches of GPT-4 were separate products stitched together under shared names. OpenAI announced the shutdown on 26 September 2025 and cut access on 26 March 2026, in a batch that also took gpt-4-0314 and gpt-4-1106-preview, recommending GPT-5 or the GPT-4.1 family as replacements. As of August 2026 the model card is still published in the catalogue of available models, complete with prices and rate limits, five months after the model stopped answering.

▸Videos

No videos yet.

▸Reviews