Grok 4.5 Slashes CursorBench Coding Costs
xAI's newly released Grok 4.5 model scored 66.7% on the CursorBench evaluation suite, trailing the leading Fable 5 Max by a small margin. Crucially, Grok 4.5 achieved this near-top performance at just $1.51 per task compared to Fable's $17.32.
Grok 4.5's massive cost reduction for near-top tier coding performance represents a significant step toward viable, production-ready AI software engineering agents.
* Grok 4.5 achieves 66.7% accuracy on CursorBench, trailing the leading Fable 5 Max by less than 4%.
* At $1.51 per task versus $17.32, Grok 4.5 delivers roughly 90% cheaper execution.
* The benchmark hints that Grok 4.5 has a specialized advantage on CursorBench, likely due to it being trained on Cursor data.
DISCOVERED
1h ago
2026-07-08
PUBLISHED
2h ago
2026-07-08
RELEVANCE
AUTHOR
bridgemindai