xAI · released
Grok 4.5
Opus-class agentic scores at a third of the price.
Overview
xAI's July release competes on agentic evals rather than academic ones. It is the cheapest model in this list that posts a serious Terminal-Bench result, and the value case is the whole pitch.
Solthar DNA rating, out of 5
- reasoning: 4
- coding: 4.5
- writing: 3.5
- vision: 3.5
- speed: 4.5
- value: 5
- context: 4
- tools: 4.5
Verdict
The value pick for agentic and coding work if you can live without published academic scores.