ctxdexSynthetic model data for learning

Benchmark family

Berkeley Function Calling Leaderboard (BFCL)

A family groups the versions of one benchmark. It is not evaluated and has no applicable models of its own; each of its 4 definitions is resolved separately.

Agentic InteractionLive EvaluationMulti Turn AgentTool Use

Definitions

Compare versions →

Newest first, following the recorded successor chain. Each definition is resolved against models separately.

  1. BenchmarkBFCL V4 AgenticV4
    ActiveNewest in family
    Successor of BFCL V3
    Scale
    26 counts
    Primary metric
    Accuracy
    Applicable models
    0
  2. BenchmarkBFCL V3V3
    Superseded
    Successor of BFCL V2 Live
    Scale
    200 counts
    Primary metric
    Accuracy
    Applicable models
    0
  3. BenchmarkBFCL V2 LiveV2
    Superseded
    Successor of BFCL V1
    Scale
    875 counts
    Primary metric
    Accuracy
    Applicable models
    0
  4. BenchmarkBFCL V1V1
    Superseded
    Scale
    200 counts
    Primary metric
    Accuracy
    Applicable models
    51

Family methodology

No family-level methodology recorded.

Across definitions

Tasks
Parallel Tool Invocation, Tool Relevance Detection, Tool Selection And Invocation, Stateful Multi Turn Tool Use, Memory Tool Use, Tool Call Format Robustness, Web Search Tool Use
Sources
None recorded at family level
Last reviewed
2026-09-23

Applicability does not mean evaluated; benchmark result/model score data is outside this UI contract.