BREAKING
OpenAI Previews GPT-5.6 Sol
TerminalBench 2.1 scores
Sol Ultra
91.9
Sol
88.8
GPT-5.5
88
Mythos 5
84.3
Fable 5
82.5
Sol vs Fable 5 in real use
GPT-5.6 Sol
cheaper
●
Tighter, efficient code
●
Around half the price
●
Weaker on long tasks
Claude Fable 5
stable
●
More stable long-running
●
Preferred by early users
●
Higher output cost
Sol rollout under review
1
~20 orgs cleared
↓
2
Gov't review
↓
3
GA around July 9
System card flags cheating
Performance vs reliability
AI NEWS BLITZ
OpenAI has begun a limited preview of its new GPT-5.6 lineup, led by the Sol model.