GPT-5(.1*) Mini + Roo,OpenCode: This is NOW WAY BETTER THAN Sonnet & Claude Code!

AICodeKingAbout 3 min readAug 18, 2025Watch original
THE SUMMARYAI-generated

Key Concepts:

  • GPT5 model (initial poor performance, potential selective serving of better models)
  • GPT5 Mini model (significant improvement, cost-effectiveness)
  • Reasoning (negative impact on GPT5 Mini performance in real-life usage)
  • One-shot tasks (GPT5 Mini excels in generating images and plans)
  • Cost-performance ratio (GPT5 Mini offers superior value)
  • Ninja Chat (AI platform with access to multiple models)
  • R Code, Klein, Cloud Code (platforms for interacting with AI models)

1. Initial GPT5 Model Performance and Controversy:

  • The GPT5 model, upon initial release, exhibited poor performance, contradicting some reviewers' positive experiences.
  • OpenAI was suspected of serving different models (potentially with higher parameters or lower quantization) to select reviewers and companies, leading to inconsistent results.
  • The Chat GPT variant of GPT5 was reportedly even worse than the API endpoints.
  • Cursor, initially promising, degraded in performance after a short period.
  • The GPT5 model's performance remained consistently poor and slower than before.

2. The Rise of GPT5 Mini:

  • The GPT5 Mini model experienced a significant improvement in performance within a few days of the initial GPT5 release.
  • The improvement was described as being on par with a 0.1 version increase in Anthropic models.
  • The speaker switched from using the Sonnet model to GPT5 Mini due to its improved performance.
  • Multiple users on Twitter reported similar improvements in specific models.

3. Reasoning and its Impact on GPT5 Mini:

  • Enabling "reasoning" in GPT5 Mini negatively impacts its performance in real-life usage.
  • Reasoning leads to slower processing, poor instruction following, token wastage, and subpar code generation.
  • The speaker recommends using GPT5 Mini without reasoning for general tasks.
  • The reasoning variant is suitable for one-shot tasks that don't require extensive editing or for planning purposes.

4. GPT5 Mini's Strengths: One-Shot Tasks and Cost-Effectiveness:

  • GPT5 Mini excels in generating images and plans, surpassing the quality of Sonnet and Opus models at a fraction of the cost.
  • Examples include generating floor plans, Pokeball images, Minecraft and Kandinsky style images, and butterfly images with detailed environments.
  • The model's cost is $0.25 for input and $0.02 for output, making it exceptionally cost-effective.

5. Coding Performance Comparison:

  • The speaker compared the coding performance of GPT5 High, GPT5 Mini (with low reasoning), and Sonnet using the same prompts.
  • GPT5 Mini, despite some initial issues with R Code, ultimately produced a satisfactory result for a relatively complex project at a cost of $0.7.
  • GPT5 High, while eventually fixing initial errors, was more expensive and produced less impressive results.
  • Sonnet generated excellent results but was significantly more expensive than GPT5 Mini.
  • GPT5 Mini is considered faster than Sonnet.

6. Ninja Chat Platform:

  • Ninja Chat is an all-in-one AI platform offering access to models like GPT40, Claude for Sonnet, and Gemini 2.5 Pro for $11 per month.
  • Features include an AI playground for comparing responses from different models and a mind map generator.
  • The basic plan includes 1,000 messages, 30 images, and 5 videos monthly.
  • Discount codes "king25" (25% off any plan) and "king40yearly" (40% off annual subscriptions) are provided.

7. Platform-Specific Performance:

  • GPT5 Mini works exceptionally well with Cloud Code and Open Code.
  • Cloud Code is optimized for non-Anthropic models, making it a good choice for GPT5 Mini.
  • Klein and Rue are heavily optimized for Claude models.

8. Conclusion:

  • The GPT5 Mini model has significantly improved and offers a superior cost-performance ratio compared to other models like Sonnet and GPT5 High.
  • Disabling reasoning is recommended for optimal performance in real-life usage.
  • GPT5 Mini is particularly well-suited for one-shot tasks and coding projects.
  • The speaker recommends using GPT5 Mini and encourages viewers to share their experiences in the comments.

AI summaries can miss context or contain errors. Check important details against the original video.

Go a little deeper.

Have a question about this video? Load its transcript to open the video chat.