#1
GPT-5.6 SolNew
gpt-5.6-solHelpfulness
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 72 tok/s
Release Date
July 9, 2026
Lab
OpenAI
Type
Proprietary
Context Size
1.1M
Max Output Tokens
128K
Cost per 1 million tokens
$5.00 / $30.00
Model Inputs*
Text, Images
Model Outputs*
Text
Tool Calling*
Enabled
Overall Assistant Score
An average score combining the 5 main categories.
95.50 pts
Rank #1
100th Percentile
0
Novice
33
Capable
66
Proficient
100
Expert
GPT-5.6 Sol is OpenAI's flagship everyday assistant for finishing real work—computer use, long tool chains, and messy multi-step help. Our tests put it at or near the top overall: unmatched helpfulness, instruction following, and comprehension, with strong natural empathy. Creative writing still pays the reasoning tax—compressed "caveman" internal scratchpads keep prose flatter than Claude Fable or the old GPT-4.5 peak—but it remains the clearest #1/#2 pick when you want an assistant that actually completes the job. Scored at cost-efficient medium/high reasoning.
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
OpenAI Models
Overall Score · Same provider
GPT Family
Overall Score · Same model family
Closest Rivals
Overall Score · Nearest by overall rank