The Race to $-Per-Token: Coding Models Get Fast & Frugal
xAI launched Grok Code Fast 1 in August 2025, offering 256k context and low pricing at $0.20 per million input tokens. Meta’s Code Llama remains free and open-source, with its 70B model rivaling proprietary benchmarks and no token fees. GitHub Copilot charges $10–$39 monthly, integrating multiple AI models for code completion. Code Llama can run locally, while Grok and Copilot require cloud access.