SAHARA - Text Generation Cluster: leaderboard

Metric: Cluster score (mean of task scores). Source: huggingface.co. 47 models tracked.

Top models

#ModelScore
1Gemini 3 Pro (Preview)18.24
2Gemini 2.5 Pro15.66
3GPT-5.6 Luna13.54
4Qwen 3.6 35B A3B12.42
5Claude Sonnet 412.04
6Qwen 3.6 27B11.96
7Qwen 3.5 27B11.73
8Llama 4 Scout Instruct11.45
9Gemini 2.5 Flash11.18
10Llama 3.1 70B Instruct11.15
11GPT-511.1
12Qwen 3.5 4B11.08
13GPT-4.111.06
14DeepSeek R1 Distill Llama 70B10.91
15Command-R+ (08-2024)10.85

Interactive version: theaggregate.ai/benchmark?slug=sahara-text-generation-cluster · How It Works · Data refreshed daily, snapshot 2026-09-19.