DeepSeek V3.2
DeepSeek · DeepSeek family · released 1 Dec 2025 · results from L2-Bench 1.0.0
DeepSeek V3.2
80.1
L2-Bench overall score
Rank 5 of 9 on L2-Bench 1.0.0 · $0.0012 per task · tested 1 May 2026
L2-Bench Score by Competency
Average score across tasks in each competency. Higher is better. Rank is among all 9 models on L2-Bench 1.0.0.
Every run
Each run is a fixed record: its score and cost never change after it is published. When a setting was tested more than once on a version, the latest run is the one on the dashboard.
| Tested | Version | Score | Cost per task | Price used | Status |
|---|---|---|---|---|---|
| 1 May 2026 | L2-Bench 1.0.0 | 80.1 | $0.0012 | price from 1 Dec 2025 | On dashboard |
Price
| From | Input, per million tokens | Output, per million tokens |
|---|---|---|
| 1 Dec 2025 | $0.31 | $0.93 |
Cost per task uses the price in force on the day each run was tested.