GPT-5.6 Luna vs GPT OSS 120B on HealthBench Hard
updated August 16, 2026
A gap of 0.020 separates these two on the 1,000 hardest HealthBench conversations: GPT-5.6 Luna at 0.320, GPT OSS 120B at 0.300. The table adds what each one costs at list rates.
Side by side
 GPT-5.6 Luna | GPT OSS 120B | |
|---|---|---|
| score | 0.320 | 0.300 |
| rank | 4 of 9 | 5 of 9 |
| context window | 1.1M | 131K |
| price per 1M tokens, in / out | $0.20 / $1.20 | — |
| 1,000-exchange workload | $1.24 | — |
| released | 2026-07-09 | 2025-08-05 |
| license | proprietary | open |
The workload row prices 1,000 exchanges of 2,000 input and 700 output tokens each, at the list rates current on August 16, 2026. GPT-5.6 models charge higher rates above 272K input tokens. GPT OSS models are open weights without vendor list pricing.
Reading the matchup
This is the deployment-model decision in two rows: OpenAI's cheapest API tier against OpenAI's strongest open-weights release. Luna scores 0.020 higher and costs $0.20 per million input tokens through the API; GPT OSS 120B runs under Apache 2.0 on your own hardware, which is the requirement that actually decides this pairing for health systems keeping records on-premises. On score alone they are one tier apart, not one class apart.
Which scores higher on HealthBench Hard, GPT-5.6 Luna or GPT OSS 120B?
GPT-5.6 Luna. On the 1,000-conversation set it scores 0.320 to GPT OSS 120B's 0.300, a margin of 0.020 as of August 16, 2026.
Which is cheaper to run, GPT-5.6 Luna or GPT OSS 120B?
GPT OSS 120B carries no vendor list price, which rules out a like-for-like cost comparison.
Related comparisons
- 0.300 vs 0.108
- 0.320 vs 0.259
Each model's full page: GPT-5.6 Luna and GPT OSS 120B. Every other pairing lives in the score-difference matrix.