BREAKING
xAI Ships Grok 4.5
0$
input / M tokens
0$
output / M tokens
0K
context window
0
Grok 4.5 tokens
0
Opus 4.8 max tokens
0x
efficiency gain
Benchmarks vs Rivals
DeepSWE 1.062
SWE-Bench Pro64.7
Terminal Bench83.3
Strong on Cost, Not Benchmarks
Strengths
Lowest cost per task
Agentic and coding workflows
Cursor-tuned training
Limits
Fourth on Elo at 1543
No single benchmark win
Not a universal leader
Cost Per Task Takes Center Stage
AI NEWS BLITZ
xAI has released its flagship Grok 4.5 model to the public, betting big on cost efficiency.