OpenAI Internal Models Already Match Gemini 3 Performance

• OpenAI Research Chief Mark Chen recently shared insights that shed light on where the company stands in the AI race. He pointed out that while Gemini 3 is a solid model, benchmarks only tell part of the story. His comments highlight a key reality: you can’t judge AI systems by test scores alone, since real-world performance often goes beyond what standardized tests capture.

OpenAI Research Chief, Mark Chen:

Gemini 3 is a pretty good model, but benchmarks only show part of the picture

We already have internal models performing at the same level as Gemini 3, and even better successors are planned for release soon pic.twitter.com/eF6Z3MnI70

— Haider. (@slow_developer) December 2, 2025

• What’s more interesting is that OpenAI is already running internal models that perform right alongside Gemini 3. This means the company’s development is moving faster than what gets announced publicly. By confirming they’ve reached this performance level internally, Chen signals that OpenAI is very much in the game at the highest level of AI development, especially as comparisons with Google’s Gemini 3 shape how people think about who’s leading the charge.

• Chen didn’t stop there. He made it clear that even better successors are planned for release soon. That’s a big deal because it shows OpenAI isn’t just keeping pace—they’re building multiple generations of models at once. The fact that he emphasized how benchmarks only show a slice of what’s possible suggests these upcoming releases might shine in areas like broader task handling, multimodal reasoning, or practical efficiency that current metrics don’t fully measure.

“Benchmarks only show part of the picture—our internal models already perform at Gemini 3 levels, and even better successors are coming soon,” Chen noted, underscoring the rapid pace of development.

• The real significance here is what this tells us about momentum in AI. With OpenAI gearing up to release models that meet or beat Gemini 3 capabilities, the competition between major AI labs is heating up fast. These developments shape what people expect from upcoming releases, influence where investment dollars flow, and suggest the next wave of frontier AI systems could arrive sooner than many anticipated.

My Take: Chen’s comments reveal that the public AI narrative lags behind internal progress. If OpenAI already matches Gemini 3 internally while planning stronger successors, we’re likely months away from seeing capabilities that reset expectations across the industry.

Source: Haider