# OpenAI’s GPT-4.1 Family: What You Need to Know in 7 Minutes
OpenAI has launched an impressive new suite of models – GPT-4.1, GPT-4.1 Mini, and GPT-4.1 Nano. These models represent significant advancements in AI capabilities while offering options across different performance and price points. Here’s a breakdown of what makes these models special.
## Key Highlights
– **Extended Context Window**: All models support up to 1 million tokens of context
– **Updated Knowledge**: Fresh knowledge cutoff of June 2024
– **GPT-4.5 Deprecation**: OpenAI is removing GPT-4.5 from their API to free up GPU resources
## Performance Improvements
### Benchmark Scores
– **GPT-4.1**: Scores 54.6% on Swebench verified benchmark (21.4% increase over GPT-4.0, 26.6% over GPT-4.5)
– **Instruction Following**: 38.3% score (10.5% increase over GPT-4.0)
– **Video Understanding**: New state-of-the-art result of 72% on no subtitle categories (6.7% increase over GPT-4.0)
### Intelligence Comparison
– Both GPT-4.1 and GPT-4.1 Mini show increased intelligence on MMLU benchmarks compared to GPT-4.0
– GPT-4.1 Nano is faster but with decreased intelligence, fitting into the “lower quadrant” of performance
## Coding Capabilities
– GPT-4.1 significantly outperforms GPT-4.0 in coding tasks, especially front-end development
– Human evaluators preferred GPT-4.1’s websites over GPT-4.0’s 80% of the time, finding them more functional and aesthetically pleasing
– Example showcased: GPT-4.1 creates more visually appealing flashcard apps with animations and better design elements
## Instruction Following
– GPT-4.1 follows instructions more reliably across different formats
– Scores 49.1% on hard instruction following prom