A few days ago OpenAI held its first developer conference - DevDay - and our team followed the event, held in San Francisco, closely.
We dug deep into the updates, talked with OpenAI support, and put together our first takeaways:
- The 16K model is now the default model. Its context is also significantly cheaper now — the price has dropped almost 3x compared to before, and it's even cheaper than the regular GPT-3.5 Turbo 4K used to cost. The model is already live on our site.
- GPT-4 is getting a 128K context window with the ability to analyze images (GPT-4V), but for now the model is only available in "preview" format and can't yet handle requests from all our users, so we're expecting it further down the line. GPT-4 pricing will also be cut in half once the GPT-4 128K model leaves "preview" mode.
- The OpenAI team has made it easier to build plugins and integrate external services into the LLM. This functionality is already available to our business clients via API, and we hope to find a use for it for regular users down the road too!
- DALL-E 3 is out! We're considering adding it to our toolkit (we'd appreciate your feedback and comments on whether you need this feature, especially if you're already using Midjourney)
On top of the great news from OpenAI, Midjourney also rolled out new functionality - a 4x increase in file size. We're already testing this feature and it will show up in our app soon!