ModelDial / Measured model comparisons
GPT-6.1 Sol vs GPT-6 Sol
Compare the two Sol generations at their highest-scoring tested overall efforts, and see how score deltas trade off against evaluation time and reference cost.
See the axis results below. Missing overall scores or different frontend scoring versions prevent an overall difference claim.
Scores, time and cost
| Metric | GPT-6.1 Solmax | GPT-6 Solxhigh |
|---|---|---|
| Overall | 93.80 | 82.70 |
| Backend engineering | 95 | 80 |
| Frontend interaction | 96 | 94 |
| Knowledge & reasoning | 90 | 75 |
| Evaluation time | 62m 9s | 39m 22s |
| Reference cost | $1.69 | $1.10 |
Each column uses all axis scores from one configuration, without mixing efforts. Costs exclude subscriptions and failed retries; missing values are shown as “—”.
Frontend scoring versions may differ. Open the full comparison to inspect the grading basis.
How to read these results
What should I check before upgrading from GPT-6 Sol to GPT-6.1 Sol?
First verify whether the capability dimension your work relies on improved, then compare evaluation time and reference cost. This page defaults to each model’s best overall effort. If you run a fixed effort in production, select that same effort in the full comparison to separate generation changes from effort changes.
This page uses ModelDial’s published evaluations, not official vendor rankings. Model detail pages provide the axis results, scoring versions and tested configurations.