In the world of artificial intelligence, Gemini 2.5 Pro has hit some impressive milestones by jointly completing Pokémon Blue. This extraordinary success story has the whole tech community buzzing. Example 1 Joel Z, a 30-year-old software engineer. He conceptualized the Gemini Plays Pokémon livestream and was instrumental in the execution of this achievement. Joel Z has been instrumental in guiding the AI’s progress through his Twitch channel, where viewers can witness Gemini tackle various challenges in the game.
As an open-source project Gemini Plays Pokémon is still very much in active development, with its framework continuing to develop to better serve its gameplay. As part of this initiative, Joel Z has emphasized the importance of not viewing the livestream as a definitive measure of an AI’s capabilities in gaming. He said that under human guidance, Gemini completed Pokémon Blue. This support improved its logic and analytical skills.
“My interventions improve Gemini’s overall decision-making and reasoning abilities,” – Joel Z
To mark this milestone, Sundar Pichai, CEO of Google, took to social media platform X to express his enthusiasm. Specifically, he tweeted Gemini’s progress so far and dropped a hint of exciting developments in AI tech for the video gaming industry. Logan Kilpatrick, another key figure in the project, remarked on Gemini’s achievements, stating that it “earned its 5th badge (next best model only has 3 so far, though with a different agent harness).”
Though Gemini’s success is impressive, it’s important to acknowledge the growing competitive landscape of AI gaming. Anthropic’s Claude AI models have made remarkable progress in completing the video game Pokémon Red. Their success is a result of deep contemplation and custom agent training. Now we have a whole dedicated Twitch channel for Claude Plays Pokémon. This is further testament to the growing excitement around what AI can do within the gaming industry.
Joel Z has encouraged viewers to be careful when interpreting comparative analyses between Gemini and other models such as Claude. He noted that direct comparisons are difficult because of the different tools and data that each AI is given.
“Please don’t consider this a benchmark for how well an LLM can play Pokémon. You can’t really make direct comparisons — Gemini and Claude have different tools and receive different information,” – Joel Z
The Gemini Plays Pokémon project is still ongoing and ever-changing. We’re looking forward to seeing how these changes develop. At the intersection of AI and gaming lies exciting groundwork for exploration and creativity. Gemini and Claude are leading this charge, fearlessly charging into this brave new world.
