Grok-4 Code (Sonic - Fully Tested): This FREE STEALTH Model is Grok-4 Code & It's Mediocre.

Key Concepts

• Sonic: A new AI model available on platforms like Rode, Klein, Open Code, and Cursor.
• Context Window: Sonic has a 262K context window.
• Grock: Speculated to be the provider of the Sonic model (XAI).
• Ninja Chat: An all-in-one AI platform offering access to various AI models.
• Temperature: A parameter affecting the randomness of the model's output (Sonic's is set to 0.7 in Ruode).
• Token Limit: The maximum number of tokens the model can output (Sonic's is capped at 16K in Ruode).
• Autoreasoning: The model's ability to reason and think through problems on its own.
• Multimodal: The ability of a model to process multiple types of data, such as text and images.
• Instruction Following: How well the model adheres to specific instructions given by the user.
• Hallucinations: Instances where the model generates incorrect or nonsensical information.

Sonic Model Overview

• Availability and Access: The Sonic model is available on Rode, Klein, Open Code, and Cursor. It is initially free for about 72 hours on R code without limits, requiring a R code cloud account. It's also free on Klein and Cursor.
• Provider Speculation: The model provider is unconfirmed, but the consensus leans towards Grock (XAI) due to reasoning traces, speed, context window size, and Grock's previous teases.
• Configuration: The model has a 262K context window. In Ruode, it operates with a temperature of 0.7 and a maximum token output limit of 16K.
• API Specifications: The model outputs an element called "number of sources used," which is typically associated with XAI APIs, suggesting live search is enabled.

Ninja Chat Advertisement

• Platform Description: Ninja Chat is an all-in-one AI platform that provides access to top AI models like GPT40, Claude 4, Sonnet, and Gemini 2.5 Pro for $11 per month.
• Features: It includes an AI playground for comparing responses from different models and a mind map generator.
• Pricing: The basic plan offers 1,000 messages, 30 images, and five videos monthly. Discount codes "king25" (25% off any plan) and "king40yearly" (40% off annual subscriptions) are available.

Performance Testing and Evaluation

• Testing Methodology: The model was tested raw without temperature adjustments or system prompts, using the Ruode Cloud API specifications.
• Scoring System: A scoring system out of 20 was used, evaluating code quality, functionality, and other factors.
• Coding Performance: Sonic is considered a "fine coding model" and a reasoning model that autoreasons when necessary.
• Comparison to Other Models: It is expected to be competitively priced with GPT5 Mini. If priced similarly, it would be a good budget reasoning model. However, it is not as good as Sonnet.
• Specific Task Performance:
  • Floor Plan Generation: Fails, producing a "chunky block."
  • SVG Panda: Resembles a "donkey."
  • Pokeball in 3JS: Poor, resembling an "ultra ball with dark colors."
  • Chess Autoplay: Makes random and illegal moves, failing to account for checkmate.
  • Minecraft Clone in Kandinsky Style: Good, with a great layout and randomness.
  • Butterfly in the Garden: Good but not as dense or detailed as GPT5 Mini.
  • CLI Tool in Rust: Produced many errors.
  • Blender Script for Pokeball: Fine but not great.
• Leaderboard Position: Sonic ranks seventh on the leaderboard.

Pricing and Recommendations

• Pricing Impact: The pricing will determine its overall value. If priced above GPT5 Mini, it is not recommended. If priced similarly, it is highly recommended.
• GLM 4.5: Praised as an excellent open model, consistently ranking in the top three.
• Multimodal Limitations: Sonic is not multimodal, unlike GPT5 Mini.

Provider Confirmation and Usage

• XAI/Grock Likelihood: The API schema, specifically the "numbum sources used" element, strongly suggests that Sonic is a Grock code model from XAI.
• User Feedback: Users are encouraged to test the model and share their thoughts.
• Platform Options: Users can access Sonic via Open Code (terminal experience), Ruode, Klein, or Cursor.

Instruction Following and Hallucinations

• Instruction Following Issues: The model exhibits poor instruction following, even in basic tasks.
• Hallucinations: The model has a significant amount of hallucinations.
• Auto Approve Unsuitability: Due to these issues, the model is not suitable for auto-approve workflows.
• Revised Expectations: Users should lower their expectations, as it is not as good as GPT5 Mini.

Conclusion

The Sonic model, potentially from Grock (XAI), offers a large context window and free initial access. While it shows promise as a coding and reasoning model, its performance varies across tasks, and it suffers from instruction following issues and hallucinations. Its ultimate value hinges on its pricing relative to competitors like GPT5 Mini. Further testing and user feedback are encouraged to determine its real-world applicability.