Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Conversational Tool-use on τ²-Bench (Domain Scores)

75.5Airline Success Rate

deepseek-v4-flash

9.4626.60543.7560.895May 18, 2026May 19, 2026May 21, 2026May 22, 2026May 24, 2026May 25, 2026May 27, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.05
75.580.91009095.4
2026.05
7058.888.67683.3
2026.05
5652.615.833.221.1
2026.05
4443.8613.1633.67-
2026.05
4243.8612.2832.71-
2026.05
3849.1215.7934.3-
2026.05
3843.914.929.124.6
2026.05
3641.2310.5329.25-
2026.05
3642.1112.2830.13-
2026.05
3242.9821.9332.3-
2026.05
3029.8225.4428.42-
2026.05
2438.613.1625.25-
2026.05
2447.374.3925.25-
2026.05
1831.5817.5422.37-
2026.05
1620.1810.5315.57-
2026.05
147.0222.8114.61-
2026.05
1218.4210.5313.65-
2026.05
1216.6716.6715.11-