GPT-4.1, GPT-4.1 mini and GPT-4.1 nano
2025-04-14 · Publisher : OpenAI · Domain : Artificial Intelligence
OpenAI launches the GPT-4.1 family (GPT-4.1, mini and nano) in the API, with better performance in code and instruction following, a one-million-token context and a first nano model.
What changes
- Three models that outperform GPT-4o and GPT-4o mini across all measures, with the knowledge cutoff moved to June 2024.
- One-million-token context for all three models (versus 128,000), with no surcharge for long requests.
- GPT-4.1 mini matches or exceeds GPT-4o on several evaluations, with latency almost halved and a cost reduced by 83%.
- GPT-4.1 nano is OpenAI's fastest and cheapest model, suited to classification and autocompletion.
- GPT-4.1 exists only in the API. The retirement of GPT-4.5 Preview is announced for July 14, 2025.
### Specifications - Context 1,047,576 tokens, maximum output 32,768 tokens, text and image input, text output, with no reasoning step. - Knowledge cutoff: June 1, 2024. - API identifiers: gpt-4.1, gpt-4.1-mini, gpt-4.1-nano (snapshots 2025-04-14); fine-tuning available for 4.1 and 4.1 mini. ### Pricing Pricing per million tokens (USD); 50% discount with the Batch API, 75% discount on the cache. | Model | Input | Cached input | Output | | --- | --- | --- | --- | | gpt-4.1 | 2.00 | 0.50 | 8.00 | | gpt-4.1-mini | 0.40 | 0.10 | 1.60 | | gpt-4.1-nano | 0.10 | 0.025 | 0.40 | ### Benchmarks Scores published by the vendor. | Benchmark | Result | | --- | --- | | SWE-bench Verified (GPT-4.1) | 54.6% (GPT-4o: 33.2%) | | MultiChallenge (GPT-4.1) | 38.3% | | IFEval (GPT-4.1) | 87.4% (GPT-4o: 81.0%) | | Video-MME, long videos without subtitles | 72.0% | | Graphwalks | 61.7% | | GPT-4.1 nano: MMLU / GPQA / Aider polyglot | 80.1% / 50.3% / 9.8% | ### Key takeaways The family targets agents and the analysis of large document volumes at a controlled cost. GPT-4.1 mini later arrived in ChatGPT in place of GPT-4o mini (May 14, 2025).
Links
Tags : LLM, OpenAI, GPT, GPT-4.1, Contexte long, API