GLM Coding Plan V/S Claude Code & Codex: IS THIS THE BEST & CHEAPEST Plan for AI Coding!

AICodeKingAbout 4 min readSep 7, 2025Watch original
THE SUMMARYAI-generated

Key Concepts:

  • GLM Coding Plan (GLM 4.5): A new coding plan from ZAI, positioned as a competitor to Claude Pro/Max and OpenAI's Plus/Pro plans.
  • Claude Code: Anthropic's coding assistant.
  • Codeex: OpenAI's coding assistant (likely referring to the coding capabilities within ChatGPT).
  • Prompt: A single user input or instruction given to the AI model.
  • Agentic Test: A benchmark designed to evaluate the performance of coding agents on complex tasks.
  • TMDB API: A popular API for accessing movie data.
  • Bubble Tea: A Go-based terminal UI framework.
  • GDScript: Godot Engine's scripting language.

1. Pricing and Availability of GLM Coding Plan

  • The GLM coding plan has two tiers: $3 and $15.
  • It's exclusively available through the Claude Code API with specific configurations.
  • The $3 plan offers approximately 120 prompts per 5-hour cycle.
  • The $15 plan provides around 600 prompts per 5-hour cycle.
  • The prompt limit is based on the number of prompts, not the number of requests made by the model in response to a prompt.

2. Configuration Process for GLM Coding Plan

  • Subscribe to the GLM coding plan.
  • Configure the API key and base URL.
  • Export the Anthropic base URL and API key obtained from the API platform.
  • Create a settings file to set the Anthropic model to GLM 4.5.

3. Comparison with Codeex and Claude Code Pricing

  • ChatGPT's $20 plan is a contender, offering good limits, potentially similar to the GLM 4.5 $3 plan. However, the limits may change as more users subscribe.
  • Claude Code's $20 plan is considered less usable, with lower limits than the GLM $3 plan.
  • Claude Code's $200 plan is a popular option.

4. Agentic Testing Methodology

  • The speaker built an agentic test to benchmark coding agents.
  • The test involves a series of questions designed to assess the capabilities of the agents.
  • The speaker intends to use this benchmark for future agent evaluations.

5. Test Case 1: Movie Tracker App

  • Task: Build a movie tracker app using the TMDB API with a good design.
  • Claude Code: Produced a decent app with the required pages in one shot.
  • Codeex (GPT-4): Failed, resulting in a lackluster UI with default Expo app leftovers.
  • GLM 4.5: Performed exceptionally well, following instructions and creating a functional app with a detail page, reviews, and a calendar. It had minor margin issues but was almost flawless and better than Codeex, and even slightly better than Claude Code.

6. Test Case 2: Go-Based Terminal with Bubble Tea

  • Task: Create a Go-based terminal application using the Bubble Tea framework.
  • Claude Code: Failed.
  • Codeex: Failed.
  • GLM 4.5: Succeeded, building the application in one shot with very few issues.

7. Test Case 3: Godot 3D FPS Shooter Game

  • Task: Add a step calculator, walking target setting, and health effects to a Godot-based 3D FPS shooter game.
  • Claude Code: Excelled, implementing all requested features in one shot.
  • Codeex: Failed due to poor GDScript knowledge, resulting in broken and unfixable code.
  • GLM 4.5: Failed, potentially due to issues with the model's knowledge, resulting in deleted scene files and a non-booting game.

8. Test Case 4: Open Code Repo with SVG Creation Wizard

  • Task: Add a new slash command to an existing code repository that opens an SVG creation wizard.
  • Claude Code: Failed.
  • Codeex: Failed.
  • GLM 4.5: Failed.

9. Overall Assessment and Recommendation

  • Codeex is considered a poor option.
  • Claude Code is good but expensive.
  • GLM 4.5 is the best value option currently available.
  • The speaker recommends the $3 GLM 4.5 plan.
  • GLM 4.5 is considered to be almost 90% as good as Claude Sonnet and even better in some cases.

10. Ninja Chat Advertisement

  • Ninja Chat is an all-in-one AI platform with access to models like GPT-4o, Claude 3 Sonnet, and Gemini 1.5 Pro.
  • It offers features like an AI playground for comparing model responses and a mind map generator.
  • Pricing starts at $11 per month for 1,000 messages, 30 images, and 5 videos.
  • Discount codes: "king25" for 25% off any plan and "king40yearly" for 40% off annual subscriptions.

11. Conclusion

The GLM 4.5 coding plan, particularly the $3 option, offers excellent value and performance compared to Codeex and Claude Code. While it may not excel in every task, its capabilities in areas like web application development and terminal application creation make it a compelling choice for developers seeking a cost-effective AI coding assistant. The speaker is switching to GLM 4.5 and recommends others to consider it as well.

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.