OpenAI: healthcare benchmark results

19 models · 30 results · updated August 16, 2026

Everything the index holds for OpenAI, gathered in one place: GPT-5.6 Sol, GPT-5.5, GPT-5.4, GPT-5.6 Terra, GPT-5.6 Sol (August), Codex (GPT-5.6-sol), GPT-5.2-High, GPT-5.6 Luna, GPT-5.4 mini, GPT-5.4 (2026-03-05), Codex (GPT 5.5), GPT 5.1, GPT-5.4 (high reasoning), GPT-5.4 (low reasoning), GPT-4.1, GPT-4o, GPT-5.4 (computer-use agent), GPT-5.5 Instant (June Update), GPT-5.6 Luna (August). The lab holds first place on MAST (Medical AI Superintelligence Test), PhysicianBench, EHR-Complex, OpenAI Dynamic Mental Health Evaluations. Scores sit on each benchmark's own scale and never compare across rows from different boards.

Every result

model	benchmark	score	position
GPT-5.6 Sol	HealthBench Professional	0.605	2 of 9
GPT-5.6 Sol	HealthBench Hard	0.331	2 of 9
GPT-5.6 Sol	HealthBench	57.0	4 of 8
GPT-5.6 Sol	MAST (Medical AI Superintelligence Test)	60.2%	1 of 11
GPT-5.6 Sol	First, Do NOHARM (v2)	70.1%	3 of 19
GPT-5.5	HealthBench	56.5	6 of 8
GPT-5.5	First, Do NOHARM (v2)	70.0%	4 of 19
GPT-5.5	PhysicianBench	46.3 ± 1.2	1 of 13
GPT-5.4	MedXpertQA (MM)	77.1%	4 of 8
GPT-5.4	PhysicianBench	27.7 ± 1.5	4 of 13
GPT-5.4	WHBench	66.8%	3 of 22
GPT-5.6 Terra	HealthBench Hard	0.327	3 of 9
GPT-5.6 Terra	HealthBench	57.0	5 of 8
GPT-5.6 Sol (August)	HealthBench	55.0	8 of 8
GPT-5.6 Sol (August)	OpenAI Dynamic Mental Health Evaluations	0.981	2 of 3
Codex (GPT-5.6-sol)	HealthAgentBench	45%	2 of 12
Codex (GPT-5.6-sol)	CHI-Bench	25.3%	7 of 45
GPT-5.2-High	HealthBench	63.3	3 of 8
GPT-5.6 Luna	HealthBench	55.8	7 of 8
GPT-5.4 mini	MedHELM	0.552	4 of 11
GPT-5.4 (2026-03-05)	MedHELM	0.538	5 of 11
Codex (GPT 5.5)	HealthAgentBench	42%	3 of 12
GPT 5.1	MedScribe (Vals AI)	88.09%	5 of 84
GPT-5.4 (high reasoning)	EHR-Complex	0.65	1 of 12
GPT-5.4 (low reasoning)	EHR-Complex	0.58	5 of 12
GPT-4.1	WHBench	51.8%	5 of 22
GPT-4o	WHBench	44.6%	6 of 22
GPT-5.4 (computer-use agent)	HealthAdminBench	26.7%	2 of 5
GPT-5.5 Instant (June Update)	OpenAI Dynamic Mental Health Evaluations	0.991	1 of 3
GPT-5.6 Luna (August)	OpenAI Dynamic Mental Health Evaluations	0.977	3 of 3

Which healthcare benchmarks does OpenAI appear on?

As of August 16, 2026, OpenAI models hold 30 current results across 15 tracked benchmarks, through GPT-5.6 Sol, GPT-5.5, GPT-5.4, GPT-5.6 Terra, GPT-5.6 Sol (August), Codex (GPT-5.6-sol), GPT-5.2-High, GPT-5.6 Luna, GPT-5.4 mini, GPT-5.4 (2026-03-05), Codex (GPT 5.5), GPT 5.1, GPT-5.4 (high reasoning), GPT-5.4 (low reasoning), GPT-4.1, GPT-4o, GPT-5.4 (computer-use agent), GPT-5.5 Instant (June Update), GPT-5.6 Luna (August).

Where does OpenAI lead?

OpenAI models hold first place on MAST (Medical AI Superintelligence Test) (GPT-5.6 Sol, 60.2%); PhysicianBench (GPT-5.5, 46.3 ± 1.2); EHR-Complex (GPT-5.4 (high reasoning), 0.65); OpenAI Dynamic Mental Health Evaluations (GPT-5.5 Instant (June Update), 0.991).

Other labs with pages: Anthropic, Google, Moonshot AI, Alibaba, Meta, xAI. The full field is on the index.