Benchmarks/Grok 4.5
SpaceXAI / Proprietary

Grok 4.5

A frontier-adjacent model with a strong coding-agent result.

INTELLIGENCE54index
CODING76%
AGENTIC1543Elo
SPEEDNOT MEASURED
FIT / INTERPRETATION

Where the evidence points.

  • Coding agents
  • Agentic tasks
  • Reasoning
TECHNICAL CONTEXT

Score needs conditions.

Context
Not disclosed
Input / 1M
Not recorded
Output / 1M
Not recorded
Access
Proprietary
EVIDENCE TRAIL

Original measurements.

Grok 4.5 evaluationOPEN SOURCE ↗