L-Eval & LongBench (v1) ↗73.11 points
GPT-4-32k · L-Eval & LongBench (v1) · closed-ended average · exam evaluation ↗
OpenAI / MODEL EVIDENCE
GPT-4-32k
2 benchmarks · 4 observations · 2 sources
Scores at a glance
2 benchmarks · large dot: GPT-4-32k; small dots: closest models where reported
τ-Bench ↗56.5%
gpt-4-32k · τ-Bench · Retail · pass^1 · Original tau-bench · function-calling agents ↗
2 more reported results
33% · gpt-4-32k · τ-Bench · Airline · pass^1 · Original tau-bench · function-calling agents · $\tau$-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains ↗
44.8% · gpt-4-32k · τ-Bench · Domain average · pass^1 · Original tau-bench · function-calling agents · $\tau$-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains ↗
Closest models by score
Similar positions on shared benchmark leaderboards. Smaller gaps are closer.
There are not enough shared scored settings to identify close comparisons reliably.
Sources
Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.
- L-Eval: Instituting Standardized Evaluation for Long Context Language Models — published result table ↗
All 8 authors
Chenxin An, Shansan Gong, Ming Zhong, Xingjian Zhao, Mukai Li, Jun Zhang, Lingpeng Kong, Xipeng Qiu
- $\tau$-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains — published result table ↗
Names used in source tables
GPT-4-32k · gpt-4-32k