SpaceXAI releases Grok 4.5 for coding and agentic knowledge work
SpaceXAI released Grok 4.5 through its API, Grok Build, and Cursor, reporting stronger coding, engineering, and long-running agentic performance. The launch also priced broad developer access and positioned the exact version as the company's most capable model.
TOWARD DOOM64confidence 83/100
Why it moved the index
The dated primary source reports a version-specific model with strong coding and engineering results, multi-hour agentic training rollouts, office-tool use, API access, and broad availability through Cursor and Grok Build. Benchmark evidence is largely developer-reported and the page provides limited independent safety evidence, so confidence is high but not maximal.
AUDIT TRAIL
Assessment history
- R1Toward 64 · confidence 83
Historical gap fill from a dated primary launch source; no Grok 4.5 story or model existed in the durable context.
11 Aug 2026