πŸ’Έ Vibe Code Cheaply With Code Puppy

A practical guide to accessing good models without breaking the bank. These are the setups I actually use.

πŸ“‹ TL;DR - The Budget Breakdown

Antigravity
$6.49/mo
Minimax
$2/mo
Z.AI
$3/mo

Total: ~$11.49/month

Access to Claude Opus 4.5, Gemini 3 Pro, GLM-4.7, and Minimax M2.1
+ GPT-5.2 if you already have ChatGPT Plus!

πŸ‡¨πŸ‡³

Minimax M2.1

$2/month
Minimax M2.1 229B params!

Chinese model providers are crushing it right now. Minimax M2.1 is a very very small model weighing in at only 229 billion parameters. How well it performs given its size is actually astonishing.

The $2/month plan at platform.minimax.io gives you introductory access with a surprisingly good amount of usage!

πŸš€ Setup:

  1. Sign up at platform.minimax.io and get an API key
  2. Run /add_model in Code Puppy
  3. Navigate to page 3 and find the Minimax provider (not the China one!)
  4. Select Minimax 2.1
  5. Paste your API key when prompted
  6. Use /model minimax-Minimax-m2.1 to switch to it
Minimax provider in Code Puppy
Finding Minimax in /add_model
Minimax M2.1 model selection
Selecting Minimax M2.1
Minimax subscription pricing
Minimax pricing - $2/month gets you started!
⚑

Z.AI (GLM-4.7)

$9/quarter ($3/month!)
GLM-4.7

Another Chinese model provider bringing the heat! πŸ”₯ The ZAI Coding plan is really goodβ€”for $9 per quarter (that's just $3/month!) you get a similar amount of usage as Minimax, but for a model that's a bit better.

I think it's nice to have both Minimax and Z.AI. They're super cheap and you can set them up in a multi-agent configuration for diverse perspectives!

πŸš€ Setup:

  1. Sign up at Z.AI and get the Coding subscription
  2. Get your API key from the dashboard
  3. Set it in Code Puppy: /set ZAI_API_KEY = ...
  4. Switch models with /model zai-glm-4.7-coding
Z.AI Coding subscription
Z.AI Coding plan - $9/quarter is a steal!
🎁

Already Have ChatGPT Plus?

You're already paying!
gpt-5.2 gpt-5.2-codex

A lot of people already have ChatGPT Plus for $20/month and don't realize they can use it for vibe coding! If that's you, you've got extra usage you didn't even know you had. Sure, you could use Codex CLI, but Code Puppy has way more featuresβ€”sub-agents, the ability to mix other providers, and that sweet multi-model configuration.

πŸš€ Setup in 30 seconds:

  1. Run /chatgpt-auth in Code Puppy
  2. Sign in with your OpenAI account (browser opens automatically)
  3. Boom! You now have access to gpt-5.2 and gpt-5.2-codex
  4. Use /model gpt-5.2 to switch to it
πŸ’‘ Pro Tip: Tune Your Reasoning!

Don't forget to play with the /model_settings command to set reasoning levels for these models! Lower reasoning for quick tasks, higher for complex architecture decisions.

🎭 Multi-Agent Magic: Diverse Critics FTW

Here's the real power move: use /pin_model <agent> <model> to assign different models to different agents. Having a diverse set of LLMs as critics is incredibly valuableβ€”each model has different strengths, biases, and blind spots.

πŸ”
Security Reviewer
Claude Opus
πŸ“
Code Reviewer
GLM-4.7
πŸ§ͺ
QA Expert
Minimax M2.1
πŸ—οΈ
Architect
gpt-5.2
πŸ’‘ Example Config

/pin_model security-auditor antigravity-claude-opus-4-5-thinking-high
/pin_model code-reviewer zai-glm-4.7
/pin_model qa-kitten minimax-Minimax-m2.1

🐢 That's the Vibe Code Way!

Code Puppy gives you the flexibility to mix and match providers, use OAuth where available, and get maximum bang for your buck. No vendor lock-in, no bloated IDEs, just pure terminal-based vibe coding with the best models available.

πŸš€ Get Started with Code Puppy