⚠ Usage-based units differ across vendors — not directly comparable.
Tokens (per 1M), requests (per call), and ACUs (agent compute minutes) measure fundamentally different things.
Click any price block to visit official pricing.
Platform plansP: Claude Code, OpenAI Codex, and Mistral Vibe subscription prices include their full platforms (Claude.ai, ChatGPT, Le Chat) — not just the coding tool. API/usage plans are standalone.
← scroll horizontally →
Key observations
- GitHub Copilot Pro ($10/mo) is the cheapest paid individual plan. Mistral Vibe ($14.99/mo) is the only EU-headquartered vendor — critical for data sovereignty.
- Team seat costs range 10× from $19 (Copilot Business) to $200 (Augment Max).
- Usage-based is apples to oranges: Copilot charges per request ($0.04), Anthropic/OpenAI/Mistral per token, Devin per ACU, Replit per effort-based Agent task. No standard metering unit exists.
- Mistral's Devstral 2 output tokens ($2.00/M) are ~12× cheaper than Claude Opus 4.6 output ($25/M).
- Platform bundlingP: Claude Code, Codex, and Vibe subscription plans include their full platforms (Claude.ai, ChatGPT, Le Chat). Their individual prices aren't directly comparable to standalone tools like Cursor or Windsurf.
- On-prem enterprise: Windsurf (FedRAMP), Mistral (EU data, fine-tune), and Cline (self-hosted by design, BYOK).