BFCL
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
BFCL Simple Python
0.938Accuracy
20
BFCL v3 (test)
31Base Score
19
BFCL V3
79.5Avg@4
19
BFCL multiple live (300 tasks model)
97.7Accuracy
18
BFCL
27.5Multi-Turn Success Rate
18
BFCL v2
88.24Non-Live AST Accuracy
18
BFCL (Berkeley Function Calling Leaderboard)
85.3BFCL Score
18
BFCL Extended Setting
85.81Non-Live Score
18
BFCL Standard Setting
86.46Non-Live Accuracy
18
BFCL Multiturn (OOD) v3 (test)
48Base Rate
18
BFCL
4.2Energy (Wh)
18
BFCL
5,093Throughput
18
BFCL
86.67Accuracy (%)
17
BFCL
90.65Non-Live Success Rate
17
BFCL Multi-Turn v4 (test)
46.75Overall Acc
17
BFCL held-out V4 (test)
52.4Average Score
16
BFCL Live
79.01Multiple Success Rate
16
BFCL Non-live
96.67Multiple Success Rate
16
BFCL
94.1BFCL Average Score
16
BFCL
1Average Rank
16
BFCL v4
70.64pass@1
16
BFCL Multi-Turn
55.5Base Performance
15
BFCL v4 (val)
85Overall Accuracy
15
BFCL Parity
85.9Accuracy
14
BFCL
86.02Single Turn Success Rate
14