SEAL - MultiChallenge (Small Language Models): leaderboard
Scale's separate MultiChallenge table for 2B-8B open models: the same multi-turn conversations, instance-level rubrics and Gemini 2.5 Pro judge as the main MultiChallenge board, ranked among small models only.
Metric: Score. Source: scale.com. Saturation forecast: Around July 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 4B (Thinking) | 25.13 | #514 (Qwen 3.5 4B) |
| 2 | Gemma 3 4B | 22.75 | #1088 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=seal-multichallenge-small-language-models · How It Works · Data refreshed daily, snapshot 2026-11-05.