Google has introduced that builders now have entry to a 2 million context window for Gemini Professional 1.5. For comparability, GPT-4o has a 128k context window.
This context window size was first introduced at Google I/O and accessible solely via a waitlist, however now everybody has entry.
Longer context home windows can result in larger prices, so Google additionally introduced help for context caching within the Gemini API for Gemini 1.5 Professional and 1.5 Flash. This permits context to be saved to be used in later queries, which reduces prices for duties that reuse tokens throughout prompts.
Moreover, Google has introduced that code execution is now enabled for each Gemini 1.5 Professional and 1.5 Flash. This characteristic permits the mannequin to generate and run Python code after which iterate on it till the specified result’s achieved.
In response to Google, the execution sandbox isn’t related to the web, comes with a number of numerical libraries pre-installed, and payments builders based mostly on the output tokens from the mannequin.
And eventually, Gemma 2 is now accessible in Google AI Studio and Gemini 1.5 Flash tuning will likely be accessible by way of the Gemini API or Google AI Studio someday subsequent month.
You might also like…
Gemini enhancements unveiled at Google Cloud Subsequent
