GPT OSS 120B on HealthBench Hard

rank 5 of 9 · updated August 16, 2026

On the 9-model HealthBench Hard board, GPT OSS 120B holds rank 5 with a score of 0.300. OpenAI's most capable open-weights model, an Apache 2.0 mixture-of-experts release that fits on a single 80GB GPU. 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

Where it sits

Muse Spark tops the board at 0.428, which puts GPT OSS 120B 0.128 off the lead. One place up is GPT-5.6 Luna at 0.320. One place down is GPT-5.3 Chat at 0.259. Every row on the board comes out of one run, so a gap between two models is measured on the same conversations under the same grader.

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 OSS 120B vs Muse Spark
  0.300 vs 0.428 · Muse Spark by 0.128
• GPT OSS 120B vs GPT-5.6 Sol
  0.300 vs 0.331 · GPT-5.6 Sol by 0.031
• GPT OSS 120B vs GPT-5.6 Terra
  0.300 vs 0.327 · GPT-5.6 Terra by 0.027
• GPT OSS 120B vs GPT-5.6 Luna
  0.300 vs 0.320 · GPT-5.6 Luna by 0.020
• GPT OSS 120B vs GPT-5.3 Chat
  0.300 vs 0.259 · GPT OSS 120B by 0.041
• GPT OSS 120B vs GPT-5.5 Instant
  0.300 vs 0.229 · GPT OSS 120B by 0.071
• GPT OSS 120B vs GPT OSS 20B
  0.300 vs 0.108 · GPT OSS 120B by 0.192
• GPT OSS 120B vs GPT-5
  0.300 vs 0.016 · GPT OSS 120B by 0.284

How the conversations are graded is on the methodology page, and how the subset was selected is on the benchmark page. The full ranking is on the leaderboard.