Anthropic just dropped some big news at their conference in San Francisco last Thursday. They introduced two new models, Claude 4 Opus and Claude Sonnet 4, for all the tech enthusiasts out there. If you’re a paying Claude subscriber, you can get your hands on Claude 4 Opus right away. But don’t worry, free and paid users will also have access to Claude Sonnet 4.
The company claims that these new models are a step up from their predecessors, boasting improved reasoning, planning, and memory capabilities. Plus, according to Anthropic’s chief product officer Mike Krieger, Claude 4 Opus is a Pokémon-playing pro. It can go agentically (whatever that means) on Pokémon for a whole 24 hours! That’s a huge upgrade from the measly 45 minutes its previous model could manage. Not too shabby, Claude.
A few months back, Anthropic launched a Twitch stream called “Claude Plays Pokémon” to showcase Claude 3.7 Sonnet’s skills in Pokémon Red. The brains behind this operation, David Hershey, decided to study how Claude could work independently on complex tasks. He chose Pokémon Red because it’s turn-based and doesn’t require real-time reactions. Plus, it holds a special place in Hershey’s heart as the first game he ever played on his trusty old Game Boy back in ’97. Hershey’s goal is to push Claude to its limits by introducing it to new games it’s never seen before. Who knows what Claude will conquer next?
When Claude 3.7 Sonnet tackled Pokémon, it hit some roadblocks. Spending hours stuck in one city and struggling to identify nonplayer characters definitely hindered its progress. But with Claude 4 Opus, Hershey noticed a significant improvement in long-term memory and planning. Watching it navigate a complex Pokémon quest, Hershey saw the AI take two days to hone its skills before moving forward. This multistep reasoning without immediate feedback shows a new level of coherence in the model. It seems like Claude 4 Opus is on the right track to success.
Everyone is buzzing about AI agents these days, and Anthropic is no exception. They’re aiming to have Claude do hours of work for users, and it looks like they’re on the right path. With the release of Claude 4 Opus and Claude Sonnet 4, Anthropic is making strides in AI safety and reliability. Their cautious approach might just pay off in the long run as they work towards creating powerful agents for consumers. But let’s not get ahead of ourselves—AI is still a work in progress, and there’s always room for improvement. Who knows what the future holds for AI technology? The possibilities are endless.
