k2.8 preview
Compare tested reasoning efforts on private datasets. Scores and costs come from their recorded evaluations.
k2.8 preview · high
- Backend & testing
- 54
- Frontend & interaction
- 72
- Knowledge & reasoning
- 70
- Evaluation time
- 70m 58s
- Reference cost
- $0.66
Totals for all three evaluations at this configuration; cost estimated from API usage.
Overall rank 23 / 33 models · Best tested effort
Backend tested · 9/11/2026
Published summary scores use the same effort. Axes may be evaluated on different dates; a backend date refers only to backend testing.
View task coverage and scoring →What changes with more effort
Compare scores, cost and time at neighboring tested efforts.
Scoring versions differ; deltas use published scores.
- Overall score
- — 64.2 → —
- Reference cost
- — $0.66 → —
- Evaluation time
- — 70m 58s → —
Results by reasoning effort
Select two or three configurations to compare scores and costs.
Swipe to see all results →
| Model / effort | Overall | Backend & testing | Frontend & interaction | Knowledge & reasoning | Evaluation time | Reference cost | Compare |
|---|---|---|---|---|---|---|---|
| high | 64.2 | 54 | 72 | 70 | 70m 58s | $0.66 | |
| max | — | — | — | 70 | 53m 24s | $0.59 |
How to read these results
The summary uses the same configuration as the highest overall score. A complete overall score requires all three axes; missing results appear as dashes.
Data snapshot published 10/1/2026