SpaceXAI model trajectory: Grok Code Fast index August 2025, July 2026

Intelligence index growth over time

The first model, Grok Code Fast 1 from August 2025, scored 21.6 on the Grok Code Fast intelligence index. Grok 4 Fast followed one month later with 27.4, a clear step forward. Grok 4.1 Fast in November 2025 reached 30.6, showing steady improvement.

In 2026 the pace accelerated. Grok 4.20 0309 in March scored 36.5, followed by Grok 4.20 0309 v2 in April with 37. Grok 4.3 at the end of April reached 37.6. The most pronounced jump came with Grok Build 0.1 0616 in June at 39.8, while Grok 4.5 in July 2026 closed the series at 53.8.

Critical interpretation of the growth

The Grok Code Fast index shows that most SpaceXAI models improved incrementally, with the exception of the sharp leap at Grok 4.5. While differences between individual versions were often small, the overall trajectory confirms a trend of rising performance.

A laboratory benchmark does not necessarily reflect real-world deployment. The intelligence index measures a specific capability, not comprehensive practical performance. Rising numbers therefore do not automatically mean better user experience or reliability across scenarios.

Implications for practice

For organisations evaluating these models, incremental improvements may suffice for specific tasks, while radical jumps such as Grok 4.5 can unlock new possibilities. Model selection should derive from needs and expected outputs, not from benchmark figures alone.

At the same time, a cadence of seven models in eleven months signals rapid development but also possible instability or a need for frequent updates. A long-term strategy and testing under real conditions are therefore essential.