BREAKING
GPT-5.6 Sol vs Claude Fable 5
Speed vs Thoroughness
GPT-5.6 Sol
Fast
●
Finished in under half the time
●
Acted autonomously
●
Used GPT Image 2
Claude Fable 5
Thorough
●
Detailed frame analyses
●
Added extra features
●
Followed instructions closely
TerminalBench 2.1 scores
Sol Ultra
91.9
Sol
88.8
Fable 5
84.3
Opus 4.8
78.9
0
%
faster task completion
0
pts
lead on Agents' Last Exam
0
pts
lead on Coding Index
App build cost: $1 vs $19
Sol app build
1
Fable app build
19
Sol API task
16
Fable API task
63
Speed or craft: pick your agent
AI NEWS BLITZ
A head-to-head app build reveals a clear split between OpenAI's Sol and Anthropic's Fable 5.