Breaking

OpenAI Boosts GPT-5.5 Instant's Health Performance

June 18, 2026 at 14:16 EDT

Foundation Models
Healthcare

On June 18, 2026, OpenAI announced that GPT-5.5 Instant, the default model in ChatGPT, has reached parity with its higher-end "Thinking" reasoning models on health-related questions. The update is available to all users, including the free tier.

June 18 · OpenAI

ChatGPT's Default Model Now Matches Its Reasoning Models on Health

GPT-5.5 Instant reaches parity with frontier "Thinking" models on health questions — and rolls out to everyone, including the free tier. More than 230 million people ask ChatGPT health and wellness questions each week.

52.5%

fewer hallucinated claims on high-risk medical, legal & financial prompts

37.3%

fewer inaccurate claims where users flagged factual errors

260+

physicians across 60 countries & 30 specialties built the evaluations

HealthBench scores: 5.3 vs 5.5 Instant

Length-adjusted scores improved across every tier (higher is better).

49.6

51.4

HealthBench

20.2

22.9

Hard

32.9

38.4

Professional

GPT-5.3 Instant

GPT-5.5 Instant

Real-world test: 18 medical prompts

Hallucinations counted in a hands-on review (each block = 1 hallucination).

GPT-5.3

→

GPT-5.5

Better accuracy on poison control & dangerous interactions, with more concise answers.

What improved

Recognizes when urgent care may be needed
Asks for relevant context & explains uncertainty
Free for all users, physician-led evaluations

The cautions

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.