GPT-5.2 on HealthBench Hard
rank 4 of 16 · updated September 8, 2026via GPT-5.6 System Card
On the 16-model HealthBench Hard board, GPT-5.2 holds rank 4 with a score of 0.343. OpenAI's December 2025 model, kept on the board as a comparison row from the GPT-5.6 system card. HealthBench Hard tests models on the 1,000 health conversations the frontier found hardest, with each response judged against its conversation's physician-written rubric and scored between 0 and 1.
Score and API facts
| rank | 4 of 16 |
|---|---|
| score | 0.343 |
| source | GPT-5.6 System Card (Section 5.1 HealthBench, Table 6 (reported as length-adjusted score (unadjusted, mean response length in characters)), column GPT-5.2), vendor-reported |
| configuration | length-adjusted, max reasoning effort, GPT-5.6 system card Table 6 column GPT-5.2 (38.9 unadjusted, 2585 chars) |
| lab | OpenAI |
| context window | 400K tokens |
| API price per 1M tokens | $1.75 in / $14.00 out |
| license | proprietary |
| released | 2025-12-11 |
Where it sits
Muse Spark tops the board at 0.428, which puts GPT-5.2 0.085 off the lead. One place up is GPT-5 at 0.347. One place down is GPT-5.6 Sol at 0.331. Rows on this board are compiled from published documents, so a gap between two models is exact only when both numbers came from the same document under the same settings; the sources page shows which document each row came from.
What does GPT-5.2 score on HealthBench Hard?
As of September 8, 2026, GPT-5.2 scores 0.343 on HealthBench Hard, 4 of 16 models on the board. The number is read from the GPT-5.6 System Card (Section 5.1 HealthBench, Table 6 (reported as length-adjusted score (unadjusted, mean response length in characters)), column GPT-5.2), vendor-reported.
What does GPT-5.2 cost per million tokens?
OpenAI lists GPT-5.2 at $1.75 per million input tokens and $14.00 per million output tokens.
Head to head
The pairings that earned a full page are linked below; the rest of the differences live in the score-difference matrix.
- GPT-5.2 vs Muse Spark0.343 vs 0.428 · Muse Spark by 0.085
- GPT-5.2 vs GPT-6 Astra0.343 vs 0.363 · GPT-6 Astra by 0.020
- GPT-5.2 vs GPT-50.343 vs 0.347 · GPT-5 by 0.004
- GPT-5.2 vs GPT-5.6 Sol0.343 vs 0.331 · GPT-5.2 by 0.012
- GPT-5.2 vs GPT-5.6 Terra0.343 vs 0.327 · GPT-5.2 by 0.016
- GPT-5.2 vs GPT-5.6 Luna0.343 vs 0.320 · GPT-5.2 by 0.023
- GPT-5.2 vs GPT-5.50.343 vs 0.315 · GPT-5.2 by 0.028
- GPT-5.2 vs GPT-5.6 Sol (August)0.343 vs 0.314 · GPT-5.2 by 0.029
- GPT-5.2 vs GPT OSS 120B0.343 vs 0.300 · GPT-5.2 by 0.043
- GPT-5.2 vs GPT-5.40.343 vs 0.291 · GPT-5.2 by 0.052
- GPT-5.2 vs GPT-5.6 Luna (August)0.343 vs 0.287 · GPT-5.2 by 0.056
- GPT-5.2 vs GPT-5.3 Chat0.343 vs 0.259 · GPT-5.2 by 0.084
- GPT-5.2 vs GPT-5.10.343 vs 0.254 · GPT-5.2 by 0.089
- GPT-5.2 vs GPT-5.5 Instant0.343 vs 0.229 · GPT-5.2 by 0.114
- GPT-5.2 vs GPT OSS 20B0.343 vs 0.108 · GPT-5.2 by 0.235
How numbers are read from their documents is on the methodology page, the document for this row is on the sources page, and how the subset was selected is on the benchmark page. The full ranking is on the leaderboard.