Bard gets a massive brain upgrade with Google's Gemini

Google is finally bringing out the heavy artillery for its AI chatbot.

Google launched its new Gemini family of models and immediately integrated it into Bard, calling it the single biggest upgrade since the tool launched.

Gemini Pro powers up Bard today

The rollout is happening in two distinct phases, starting right now with text-based tasks.

Bard is now running on a specifically tuned version of Gemini Pro in English across more than 170 countries and territories. While multimodal inputs and regions like Europe are slated for later, text prompts get an instant boost in reasoning, coding, summarizing, and planning.

Google brought receipts, too. In industry benchmarks, Gemini Pro beat GPT-3.5 in six out of eight tests, including MMLU for general understanding and GSM8K for math reasoning.

According to Google, third-party blind raters also ranked the new Bard as the most preferred free chatbot against leading alternatives. To show off its practical chops, Google even teamed up with YouTuber Mark Rober to design a high-precision paper airplane using the bot.

Ultra and Bard Advanced coming early next year

Gemini was built in three tiers: Nano for on-device tasks, Pro for versatile scaling, and Ultra for heavy lifting across text, video, audio, image, and code.

The real flagship experience will arrive early next year with Bard Advanced, which will give users their first hands-on look at Gemini Ultra. Google noted it is currently running extensive safety evaluations and will kick off a trusted tester program before opening the doors wider.

The update also keeps Bard's classic safeguard, the Google it button, intact so users can double-check model responses against web search results.

Why it matters

For everyday users, the bar for free AI assistants just went up significantly.

If you have been relying on the free tier of ChatGPT running on GPT-3.5, trying the upgraded Bard right now is a no-brainer. Even though it is currently limited to English text prompts, getting benchmark-beating logic, math, and coding assistance for zero dollars makes it an immediate contender for daily research and brainstorming workflows.

For the broader AI landscape, this puts real competitive pressure on OpenAI. For months, Google's chatbot felt like it was playing catch-up, but tuning Pro to clearly outpace GPT-3.5 on paper gives Google a solid defensive anchor in the free market.

We think Google's two-step rollout is a pragmatic maneuver. Shipping the stable Pro model today lets them stem the bleeding of user interest, while gating the massive Ultra model behind safety evaluations avoids a rushed public blunder. That said, the real test will come early next year when Bard Advanced goes live, where it will have to prove whether Gemini Ultra can truly take on frontier models in real-world friction.