- ML News
- Posts
- Nano Banana Pro & Gemini 3 Dominate AI Benchmarks
Nano Banana Pro & Gemini 3 Dominate AI Benchmarks
ML News
November 23, 2025
This summary highlights the recent advancements and benchmark achievements of Google’s Gemini 3 model, particularly the Nano Banana Pro (Gemini 3 Pro Image) variant, and early explorations of GPT-5 in scientific applications. Nano Banana Pro has achieved top performance on the LMArena benchmark.
A new era of intelligence with Gemini 3
📝Google releases Gemini 3, its most intelligent model, marking a significant advance in AI capabilities with potential impact across various applications for ML developers.
Build with Nano Banana Pro, our Gemini 3 Pro Image model
📝Announcing Nano Banana Pro, or Gemini 3 Pro Image, which is the most advanced image generation and editing model. This release can empower developers to unlock new image-based AI applications.
Introducing Nano Banana Pro
📝Google DeepMind introduces Nano Banana Pro, a new image generation and editing model which could allow developers to create/edit images with much more ease.
Early experiments in accelerating science with GPT-5
📝OpenAI reveals cases where GPT-5 accelerates scientific progress across multiple disciplines, which represents how AI can be a valuable tool for scientific discovery.
Gemini 3 Pro Image (Nano Banana Pro) takes #1 on LMArena with a record breaking +84 point lead over SOTA
📝Reports showing the Gemini 3 Pro Image (Nano Banana Pro) has set a new record on the LMArena leaderboard. Nano Banana’s score indicates that this is an important leap in image generation capabilities and quality.