Jazzlike_Course_9895 shares initial experiences using GPT-5 in Visual Studio Code Copilot, highlighting differences from previous AI models and discussing how GPT-5’s analytic code feedback compares to alternatives like Claude 4 and GPT-4.1.

First Impressions of GPT-5 Integration in Visual Studio Code Copilot

Author: Jazzlike_Course_9895

Overview

The post discusses the availability of GPT-5 within GitHub Copilot in Visual Studio Code, offering initial impressions and early user experience feedback. The author compares GPT-5’s code summarization and improvement suggestions to GPT-4.1 and Claude 4, evaluating output quality, analytic focus, and the practical impact on coding workflows.

Availability and Setup

  • GPT-5 is now reportedly enabled for some users in Visual Studio Code’s Copilot integration.
  • Users should check and enable the feature within their Copilot settings.
  • Some subscription users (like those with Windsurf) see multiple GPT-5 variants (low, medium, high reasoning levels) with some included as free-of-credits options.

Model Comparison Table

Model Summary Style Number of Points Example Code Focus
GPT-5 Very analytic, long, less human 10 (+ subpoints) Yes Mix of code & generic advice
GPT-4.1 Short, limited analysis 6 (with fluff) Yes Partial code focus
Claude 4 Concise, code-centric 5 Five Strictly code improvements
  • GPT-5 produces the lengthiest, most analytic summaries but sometimes includes excessive generic advice that may not be directly actionable.
  • GPT-4.1’s summaries are considered the weakest, with some improvement but still lacking code-centric feedback.
  • Claude 4 is praised for “grumpy senior dev” style—direct, practical code-focused suggestions, and concise output without filler.

User Observations

  • GPT-5’s approach is seen as more academic, with detailed lists and analytic explanations, but sometimes loses the practical, human touch.
  • Claude 4 offers precise code analysis and actionable improvements, favored by the author for this use case.
  • Disagreement about GPT-5’s value at “1 premium request price”—comparable to Claude 4 with less perceived benefit.
  • Additional feedback on different subscription plans (e.g., Windsurf) and the lack of multiple GPT-5 variants for some users.
  • Speculation on model pricing and API access for cost-effectiveness.

Anecdotal Conclusion

  • The author concludes that, for now, Claude 4 better meets coding feedback needs, while GPT-5 is promising but perhaps too verbose for practical development tasks.
  • Further experimentation is planned, especially for agent usage and more complex reasoning prompts.

Community Sentiment

  • Some frustration about pricing, expectation management, and the desire for more flexible or lower-cost access to advanced models in Copilot.
  • Users note differences in reasoning quality, warmth of feedback, and realism between models.

References:

This post appeared first on “Reddit Github Copilot”. Read the entire article here