Atlas

Models

← All models

Grok Build 0.1

xAI · 2026-05-29

Grok Build 0.1 is SpaceXAI's coding model for agentic software-engineering workflows, including web development, debugging, tool use, and MCP. It is the model behind the Grok Build CLI and is also available through the xAI API.

Benchmark scores

BenchmarkScore
AA-LCR64.7
AA-Omniscience Index7.4
Artificial Analysis Agentic Index28.0
Artificial Analysis Coding Index51.5
Artificial Analysis Intelligence Index39.8
Artificial Analysis Omniscience Accuracy51.4
Artificial Analysis Omniscience Hallucination Rate90.5
BLXBench72.8
BLXBench Coding category95.4
BLXBench Cost category92.7
BLXBench Debugging category74.7
BLXBench Hallucination category70.8
BLXBench pass rate51.0
BLXBench Reasoning category72.3
BLXBench Refactoring category60.1
BLXBench Security category63.1
BLXBench Speed category97.2
BLXBench UI category65.4
BuseyBench SVG3.6
Capability151.0
ChessBench58.7
ChessBench — Fork33.3
ChessBench — Hanging Piece43.3
ChessBench — Mate in 196.7
ChessBench — Mate in 286.7
ChessBench — Pin33.3
CritPt9.1
DeepSWE v113.1
DeepSWE v1 pass@429.2
GBA Eval Overall2.4
Loading Atlas data…