THE SUMMARYAI-generated
Key Concepts:
- Gro 3 API availability
- Gro 3 vs. Gro 3 Mini models
- Reasoning capabilities and thinking tokens
- Pricing comparison with Claude and other models
- Free credits program
- Integration with Klein, Rode, and Requesty
- Ninja AI platform
- Code generation performance
- Multimodality support
1. Gro 3 API and Model Overview
- Gro 3's API is now available, including a new model.
- Two models: Gro 3 (expensive, no reasoning) and Gro 3 Mini (cheap, with reasoning).
- Gro 3 costs $3 and $15 for input and output tokens, respectively, similar to Claude.
- Gro 3 Mini costs $0.30 and $0.50 for input and output tokens, respectively.
- Gro 3 reasoning was initially in beta and used in "big brain mode," while the current reasoning on the Gro platform uses Gro 3 Mini.
2. Reasoning Capabilities and Performance
- Gro 3 Mini includes a reasoning option with low and high settings.
- The API response includes "thinking tokens," which is a positive feature.
- Gro 3 Mini outperforms Gro 3 in some cases due to its reasoning capabilities.
- Gro 3 Mini's performance is similar to 03 Mini while being cheaper.
- The speaker notes that users will likely be charged for thinking tokens.
- On the Ader leaderboards, Gro 3 Mini (high reasoning) scored 49.3% at $0.73, while Gro 3 scored 53% for $11, making the Mini model more cost-effective.
3. Pricing and Free Credits
- Gro offers $150 worth of free credits per month for users who share responses and API data.
- Users need to have spent at least $5 to enable the free credits, but those who spent $5 previously can still access it.
- The free credits are beneficial for students and those who don't heavily rely on code generation.
4. Integration with Development Tools
- Gro 3 is available on Open Router and Requesty.
- To use with Klein and Rode, upgrade to the latest versions in VS Code.
- Klein and Rode allow selecting XAI and choosing Gro 3 or Gro 3 Mini.
- Issue: Klein and Rode lack a reasoning effort slider for Gro 3 Mini, defaulting to low reasoning effort.
- Open Router also lacks an endpoint for automatic high reasoning.
- Requesty has different model names for low and high reasoning efforts (Grock 3 mini low and Grock 3 mini high).
5. Code Generation Example
- The speaker demonstrates using Gro 3 Mini to create a 3JS game of a sailing ship.
- The model quickly generates the code, and the thinking traces are available.
- The generated game looks good, and the cost is low.
6. Comparison with Other Models
- Gro 3 Mini is considered a "real deal" due to its low cost and good performance.
- It is less censored than other models.
- While not as cheap as Gemini's Flash, it offers a lucrative price point.
- The speaker plans to use it in their workflow.
- Gro 3 Mini is better at coding than Haiku, which is more expensive.
7. Requesty and API Key Usage
- Requesty allows users to input their own API keys, enabling the use of the $150 free credits.
- To use with Root Code, select the provider (X or Open Router), but reasoning effort won't work.
- Users need to switch to Requesty and choose the high or low reasoning model.
8. Multimodality Support
- Gro 3 does not currently support multimodality, but the functionality is planned for the future.
9. Ninja AI Platform
- Ninja AI is an all-in-one platform with access to top AI models like GPT40, Claude 3.7 Sonnet, and Gemini 2.0 Flash for $11 per month.
- Features include an AI playground for comparing model responses and a mind map generator.
- The basic plan includes 1,000 messages, 30 images, and five videos monthly.
- Use code "king25" for 25% off any plan or "king40yearly" for 40% off annual subscriptions.
10. Conclusion
- Gro 3 Mini is a snappy, awesome model with a lucrative price point.
- The $150 free credit is a significant advantage.
- The speaker encourages viewers to try the model and share their thoughts.
AI summaries can miss context or contain errors. Check important details against the original video.