BREAKING
GPT-5.5 Instant Matches Thinking on Health
0
M
weekly users
0
+
physicians
0
countries
0
%
fewer hallucinations
0
%
fewer inaccuracies
HealthBench scores rise
HealthBench 5.3
49.6
HealthBench 5.5
51.4
Pro 5.3
32.9
Pro 5.5
38.4
Gains, but verification still needed
Improvements
●
Free tier rollout
●
Physician-led evaluations
●
More concise answers
Cautions
●
Hallucinations halved, not gone
●
Not ward-care level
●
Human check still needed
Speed meets reasoning on health
AI NEWS BLITZ
OpenAI says its default ChatGPT model now matches its reasoning models on health questions.