THE SUMMARYAI-generated
Key Concepts
- Code Supernova: A new stealth multimodal model, suspected to be a Grock model.
- Multimodality: Supports image inputs for visual context understanding.
- Context Window: 200,000 tokens (128,000 input, 64,000 output).
- Reasoning Model: Adjustable reasoning levels (low, medium, high).
- Kylo Code: A platform used for testing the model.
- Grock: An AI model, possibly the base for Code Supernova.
- Ninja Chat: An all-in-one AI platform offering access to various AI models.
- Agentic Questions: A set of questions used to test the model's agentic capabilities.
- GLM4.5: A competing model used for comparison.
- Code Fast: A previous version of Grock's coding model.
1. Introduction of Code Supernova
- A new stealth model called Code Supernova has been launched on multiple platforms.
- The model is suspected to be a Grock model, possibly by XAI, based on internal API outputs.
- It is a multimodal model that currently supports image inputs but not video inputs.
- The model has a 200,000 token context window, with a possible split of 128,000 for input and 64,000 for output.
- It is a reasoning model with adjustable levels (low, medium, high), unlimited Kylo code usage, and no throttling.
- The model is described as fast and good at visual context understanding, interpreting screenshots, diagrams, and mockups alongside text.
2. Use Cases and Applications
- Debugging: Useful for sharing screenshots of error messages, console outputs, or unexpected UI behavior for accurate debugging help.
- Design to Code: Translates mockups, wireframes, or hand-drawn sketches into code, focusing on multimodality.
- Front-end Development: Helps understand requirements when building front-end heavy projects.
3. Testing Methodology
- The model was tested using four agentic questions within the Kylo code environment.
- Kingbench questions were not used because the model is only available on coding agents, not the API, making fair testing difficult due to system prompting.
4. Ninja Chat Advertisement
- Ninja Chat is introduced as an all-in-one AI platform for $11 per month, providing access to models like GPT40, Claude 4, Sonnet, and Gemini 2.5 Pro.
- Features include an AI playground for comparing responses and a mind map generator.
- The basic plan includes 1,000 messages, 30 images, and five videos monthly.
- Discount codes "king25" (25% off any plan) and "king40yearly" (40% off annual subscriptions) are provided.
5. Evidence for Grock Model Identification
- The model's request logs reveal "numbum sources used," an element exclusive to the Grock models API.
- This evidence strongly suggests that Code Supernova is a Grock model, possibly a larger version of Grock code.
6. Testing Results and Performance Analysis
- Movie Tracker App: The model generated a movie tracker app, but the container for movies overflowed, and it loaded all movies without limits, causing memory and API spam issues. The inner pages looked good, but the top bar was not removed as it should have been.
- Goateui Calculator: The model failed to align the elements well.
- Godo Game (FPS with Step Counter and Life Bar): The model performed well, successfully adding a step counter and a life bar affected by player jumps. The UI was usable, and it worked in one shot.
- Open Code Test (SVG Generation Modal): The model failed and glitched, repeatedly saying "let me check" before Kylo code stopped it. This glitching behavior was also observed in Grock code fast.
7. Comparative Analysis
- Claude Code Gen is considered better for UI cohesion and usability in the first try.
- GPT5 codecs are also superior.
- GLM4.5 is recommended for everyday usage and agentic contraptions due to its reliability.
8. Strengths and Weaknesses
- Strengths:
- Fast performance.
- Good at Godo game development.
- Multimodality with image input support.
- Weaknesses:
- Struggles with longer tasks and complex UI alignment.
- Prone to glitching during long generations.
- Less reliable than GLM4.5.
9. Conclusion
- Code Supernova is a promising coding-focused model, possibly a better version of Grock code with multimodality.
- It is fast and shows potential, but it needs improvements in reliability and handling longer tasks.
- GLM4.5 is currently recommended for everyday use, but Code Supernova's multimodality is a significant advantage.
- The model's performance is similar in Rue Clin and Open Code.
- It is a good first coding-focused model from a relatively new lab like Grock.
AI summaries can miss context or contain errors. Check important details against the original video.





