GPT-5.6 Sol vs GPT-5 on HealthBench Hard
updated August 16, 2026
A gap of 0.315 separates these two on the 1,000 hardest HealthBench conversations: GPT-5.6 Sol at 0.331, GPT-5 at 0.016. The table adds what each one costs at list rates.
Side by side
 GPT-5.6 Sol | GPT-5 | |
|---|---|---|
| score | 0.331 | 0.016 |
| rank | 2 of 9 | 9 of 9 |
| context window | 1.1M | 400K |
| price per 1M tokens, in / out | $5.00 / $30.00 | $1.25 / $10.00 |
| 1,000-exchange workload | $31.00 | $9.50 |
| released | 2026-07-09 | 2025-08-07 |
| license | proprietary | proprietary |
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
The widest gap of the curated pairings: 0.315 between OpenAI's current flagship tier and its August 2025 one, run under the same protocol. GPT-5's near-zero result here is a default-settings API number; OpenAI's own launch-era system card reported 0.255 for the main configuration and 0.462 with thinking enabled, graded under its own setup. Whichever configuration you trust, the 5.6 line handles these conversations at a level the 2025 flagship's defaults never reached.
Which scores higher on HealthBench Hard, GPT-5.6 Sol or GPT-5?
GPT-5.6 Sol. On the 1,000-conversation set it scores 0.331 to GPT-5's 0.016, a margin of 0.315 as of August 16, 2026.
Which is cheaper to run, GPT-5.6 Sol or GPT-5?
GPT-5. The same workload of 1,000 exchanges (2,000 input and 700 output tokens each) comes to $9.50 on it and $31.00 on the other.
Related comparisons
- 0.428 vs 0.331
- 0.331 vs 0.327
Each model's full page: GPT-5.6 Sol and GPT-5. Every other pairing lives in the score-difference matrix.