TeleResilienceBench - ORANBench: leaderboard
Metric: Correct flip rate (CFR, %): share of the 63 ORANBench instances (O-RAN knowledge, GSMA Open-Telco ot-lite) where the model, given the question, the options and the first half of a flawed reasoning trace from qwen3.5:2b that ended in a wrong answer, continues the reasoning in its thinking stream and returns the correct option; one continuation prompt, Ollama builds on one RTX 4090; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | nemotron-3-nano:4b (Ollama) | 39.7 |
| 2 | gemma4:e4b (Ollama) | 38.1 |
| 3 | gemma4:e2b (Ollama) | 36.5 |
| 4 | gemma4:26b (Ollama) | 33.3 |
| 5 | gemma4:31b (Ollama) | 31.7 |
| 6 | qwen3.5:4b (Ollama) | 28.6 |
| 7 | qwen3.5:9b (Ollama) | 23.8 |
| 8 | qwen3.5:27b (Ollama) | 23.8 |
Interactive version: theaggregate.ai/benchmark?slug=teleresiliencebench-oranbench · How It Works · Data refreshed daily, snapshot 2026-10-07.