BREAKING
xAI Releases Grok 4.5 for Coding
0T
parameters
0TPS
inference speed
0x
fewer tokens
Output Tokens vs Claude Opus 4.8
Grok 4.515954
Opus 4.867020
How Grok 4.5 Was Trained
1GB300 pre-training
2Cursor data
3RL tuning
4Grok Build rollouts
Strengths vs Open Questions
Strengths
Large codebases
App building from one prompt
Excel and Office work
Caution
Limited independent benchmarks
Still tuning via RL
Claude parity unverified
Verification Still Ahead
AI NEWS BLITZ
xAI has launched Grok 4.5, its most capable model built for coding and agentic tasks.