#11
Grok 4.5New
grok-4.5Helpfulness
Instruction
Following
Comprehension
Empathy
Creative
Writing
Helpfulness
0.0
Instruction Following
0.0
Comprehension
0.0
Empathy
0.0
Creative Writing
0.0
Speed
Avg 86 tok/s
Release Date
July 8, 2026
Lab
xAI
Type
Proprietary
Context Size
500K
Max Output Tokens
128K
Cost per 1 million tokens
$2.00 / $6.00
Model Inputs*
Text, Images
Model Outputs*
Text
Tool Calling*
Enabled
Overall Assistant Score
An average score combining the 5 main categories.
90.50 pts
Rank #11
85th Percentile
0
Novice
33
Capable
66
Proficient
100
Expert
Grok 4.5 is xAI's July 2026 flagship assistant—faster and more reliable at finishing real work than earlier Grok 4.x chat models, while keeping a direct, less-hedged personality. Our tests show clear gains in helpfulness, instruction following, and comprehension on multi-step everyday tasks. Empathy is engaging and opinionated rather than soft. Creative writing is punchy in short bursts but our tests find less sustained nuance than Claude or the old GPT-4.5 peak—agent- and coding-oriented training favors getting it done over literary voice. A strong all-around assistant when you want speed and a human-feeling chat style.
Overall Score vs Price
Price axis is logarithmic (USD per 1M input tokens) · Higher score and lower price is better
Ringed point is this model
Anthropic
DeepSeek
Google
Meta
Moonshot
OpenAI
SpaceX AI
Zhipu AI
Intelligence
Overall Score · Higher is better
Speed
Output tokens per second · Higher is better
Price
USD per 1M input tokens · Lower is better
SpaceX AI Models
Overall Score · Same provider
Grok Family
Overall Score · Same model family
Closest Rivals
Overall Score · Nearest by overall rank