BREAKING
Anthropic Releases Claude Sonnet 5
SWE-Bench Pro: 63.2%
Sonnet 4.6
58.1
Sonnet 5
63.2
0
$
Input / M tokens
0
$
Output / M tokens
0
M
Context window
Where Sonnet 5 Sits
Sonnet 5
new
●
Default on Free and Pro
●
Close to Opus 4.8 on agents
●
Cost-efficient daily driver
Opus 4.8
●
Still leads on some tasks
●
Higher price tier
Key Behavior Changes
1
Adaptive thinking on
↓
2
Manual thinking removed
↓
3
New tokenizer, ~30% more
↓
4
Cybersecurity limited
A Lift in Everyday Coding
AI NEWS BLITZ
Anthropic has launched Claude Sonnet 5, its new balanced flagship for coding and agents.