Jin Daily AI Trivia: 5 Things Sam Didn’t Tell You About GPT 4.1

Jin Daily AI Trivia: 5 Things Sam Didn’t Tell You About GPT 4.1

• OpenAI will sunset the GPU-hungry GPT-4.5 and replace it with GPT-4.1 (not enough GPU capacity for more training).

• If you’re already using GPT-4o-Mini, switching to GPT-4.1-Mini will actually cost you more. GPT-4.1-Nano is cheaper, but the performance isn’t as good. So don’t switch just yet. (Free ChatGPT users are on GPT-4o-Mini.)

• The main reason to switch to the GPT-4.1 series is the 1M token context window. The older GPT-4o model only supports 128K tokens.

• The GPT-4.1 series is designed almost specifically for coding and API usage. It follows precise instructions and is optimized for structured tasks—so no more system prompt hacks or yelling at your AI for better output.

• For multimodal tasks (image/video), GPT-4.1-Mini performs similarly—or even better—than the larger GPT-4o model. Great news if you work with multimodal content: costs have dropped from $2.30 to $0.42 per 1M tokens (blended usage). That’s an 82% cost reduction!

PS: personal test, nano is crap. Don’t use it unless u need OpenAI stuff for dirt cheap. 4.1-mini is around old Deepseek V3 but perform worse than V3-0324.

Token Input Costs (per 1M tokens):

Model Cost (USD) GPT-4.1 2.00 GPT-4.1-Mini 0.40 GPT-4.1-Nano 0.10 GPT-4o 2.50 GPT-4o-Mini 0.15 Gemini 2.5 Pro 1.25–2.50 Gemini 2.0 Flash 0.15 Claude 3.7 Sonnet 3.00 LLaMA 4 Maverick 0.50 DeepSeek R1 0.135–0.55 DeepSeek V3 0.135–0.27