xAI Releases Grok 4.5, a Coding‑Focused Frontier Model
Update your integration to use Grok 4.5 and adjust cost calculations to the new pricing.
Add Grok 4.5 to your model list and test performance on coding tasks.
Summary
xAI’s Grok 4.5, trained with Cursor, is a 1.5 T‑parameter model positioned as Opus‑class but faster, more token‑efficient, and lower cost. The model offers $2 per million input tokens and $6 per million output tokens, with a 75% discount on cache hits and a 500k context window (likely to return to 1M soon). Grok 4.5 is available in Grok Build, API, and Cursor, with day‑0 support in Hermes Agent, Portal, and OpenRouter. The release emphasizes coding and agents, targeting the market dominated by Anthropic, OpenAI, and Cursor‑style tool‑using systems. The model is 3× larger than Grok 4.3, and xAI highlighted its speed and cost efficiency as key differentiators.
The launch marks xAI’s first entry into the flagship coding‑agent tier, offering a new option for developers seeking high‑performance, cost‑effective coding agents without the need for extensive fine‑tuning.
Key changes
- Grok 4.5 is 1.5 T‑parameter, 3× larger than 4.3
- Positioned as Opus‑class but faster, token‑efficient, and lower cost
- Pricing $2 in / $6 out, 75% cache hit discount, 500k context window
- Available in Grok Build, API, and Cursor with day‑0 support in Hermes Agent
- Training with Cursor, targeting coding and agents
- xAI’s first flagship coding‑agent tier release
- Speed and cost efficiency highlighted
- Target market dominated by Anthropic, OpenAI, and Cursor