🖼️ AI Image Generation Showdown: Nano Banana Pro vs Chestnut
The latest advancements in AI image generation models are transforming creative possibilities with unprecedented speed, quality, and editing capabilities.
⚡ Speed & Performance Breakthrough
Nano Banana Pro leads the pack with lightning-fast 10-20 second generation times. While OpenAI’s Chestnut shows significant improvement over previous models, it still trails behind Google’s offering in raw speed performance.
📝 Text Accuracy Revolution
Nano Banana Pro has achieved what many considered impossible: solving the historical Achilles’ heel of AI image generation – accurate text rendering in images. This breakthrough enables perfect text integration in generated visuals.
✏️ Advanced Editing Capabilities
Nano Banana Pro offers a sophisticated editing suite with context-aware conversational editing, multi-turn refinement, and remarkable consistency across edits without identity drift – maintaining the integrity of subjects throughout modifications.
🔍 Resolution Quality Standards
Nano Banana Pro delivers stunning 4K resolution (up to 8 MP) for premium users, setting the current benchmark for image quality in the industry. This represents a significant leap forward in the detail and clarity possible with AI-generated imagery.
🗓️ Strategic Release Timing
OpenAI’s Chestnut is expected to debut alongside GPT-5.2, indicating a coordinated product launch approach. This strategy suggests OpenAI is positioning these technologies to work in concert for enhanced capabilities.
🏆 Industry Competition Heats Up
Google and OpenAI are now in a neck-to-neck race for AI image generation supremacy, with both tech giants rapidly iterating on their models. This fierce competition is driving unprecedented innovation and improvement cycles in the field.
OpenAI’s Secret ‘Chestnut’ & ‘Hazelnut’ Models: The Image-2 Leak Explained
The AI image generation war just got hotter. If you’ve been watching the leaderboards lately, you might have noticed two mysterious new challengers climbing the ranks. Codenamed "Chestnut" and "Hazelnut", these models are currently being tested in the wild on platforms like LM Arena and Design Arena.
Rumors are swirling that these are the early versions of OpenAI’s next flagship image generators, likely to be branded as Image-2 and Image-2-mini. Their mission? To take down the current reigning champion: Google’s Nano Banana Pro.
Here is everything we know about these secret models as of December 13, 2025.
🌰 What Are Chestnut and Hazelnut?
OpenAI has a history of testing new models under "bot" names before officially releasing them. "Chestnut" and "Hazelnut" fit this pattern perfectly.
- Chestnut appears to be the larger, more powerful model (possibly Image-2). Early testers report it has a deep understanding of complex prompts and world knowledge.
- Hazelnut seems to be a faster, lighter version (possibly Image-2-mini), designed for speed and efficiency without sacrificing too much quality.
These models appeared quietly on LM Arena, a popular crowdsourced benchmarking platform where users vote on which AI output is better without knowing the model names first. This "blind test" strategy helps OpenAI gather unbiased data before a public launch.
⚔️ The Rival: Google’s ‘Nano Banana Pro’
To understand why "Chestnut" is a big deal, you have to look at what it’s up against. Right now, the heavyweight champion is Google’s Nano Banana Pro (powered by Gemini 3 Pro).
Nano Banana Pro set a new standard in late 2024/early 2025 with features that creators love:
- 14-Image Consistency: It can keep a character's face the same across 14 different generated images.
- Perfect Text: It renders long paragraphs of text on signs and slides with almost zero typos.
- Integration: It’s built deeply into the Gemini ecosystem.
OpenAI’s DALL-E 3 was great, but it has started to lag behind Google’s "Banana" speed and consistency. Chestnut is OpenAI’s answer to regain the throne.
🚀 Key Features Spotted in Testing
While OpenAI hasn't released official docs yet, keen-eyed users on Design Arena have noticed some "next-gen" capabilities in the Chestnut outputs:
- Code in Images: Unlike previous models, Chestnut seems capable of rendering accurate code snippets inside the image itself (like a photo of a computer screen showing Python code).
- Hyper-Realism: The "plastic" look of DALL-E 3 is gone. Chestnut’s lighting and textures are reportedly much closer to photography.
- Improved Text: Early results show it rivals Nano Banana Pro in spelling accuracy, easily handling complex signboards and logos.
📅 When Will It Release?
There is no official release date yet. However, the appearance of these models on public testing arenas usually signals that a launch is close—likely within weeks.
Many analysts predict that Image-2 (Chestnut) will launch alongside the rumored GPT-5.2 update, creating a powerful new multimodal suite for developers and creators.
💡 What This Means for You
If you are a content creator or digital marketer, here is the bottom line:
- Don't cancel your subscriptions yet. Nano Banana Pro is still the king today.
- Watch the space. If Chestnut delivers on its rumors, we might soon have a tool that can generate perfect code screenshots and consistent characters for your YouTube thumbnails and blog posts.
- Pricing: Expect the "Hazelnut" (Mini) version to be much cheaper, making it great for high-volume automation tasks in Make.com.
Stay tuned to SoftReviewed for more updates as we track these "nuts" in the wild!
