xAI's Grok 4.5 undercuts rivals with $2/$6 coding pricing, targets agentic tasks

xAI's Grok 4.5, a coding-focused model, returned to attention this week on the strength of aggressive pricing: $2 per million input tokens and $6 per million output tokens, which xAI positions as undercutting GPT-5 and Claude by up to 80% on coding costs. The model is deliberately pitched at software engineering and autonomous agent workloads rather than consumer chat, emphasizing speed and affordability.

The strategic play is price-led displacement in the developer segment — the same wedge DeepSeek is driving from the Chinese side. With OpenAI throttling Astra usage and Anthropic's Max plans drawing complaints about value, a cheap, fast coding model has an obvious opening among cost-sensitive engineering teams and agent builders running high token volumes.

The caveat is quality: reporting notes Grok 4.5 still trails rivals like GPT-5 and Claude on some benchmarks, so the pitch is explicitly price-performance rather than frontier capability. There's also the broader xAI narrative — ongoing speculation about integrating Grok with Tesla's Full Self-Driving stack, with engineers reportedly using Grok infrastructure in FSD development and pursuing Grok voice integration in-car. That's a longer-horizon vision, not a shipped product. Watch whether the pricing actually moves developer market share, how Grok 4.5 fares on agentic-coding leaderboards against Qwen3.8-Max and Claude, and whether the Tesla integration produces anything concrete.