Scale SEAL MultiChallenge leaderboard: who is #1

Name: Scale SEAL MultiChallenge rankings
Creator: Scale AI
License: https://creativecommons.org/licenses/by/4.0/

Scale AI · Score, higher is better · 29 models · published Sep 24, 2026

Top 25 of 29 models, best variant per model as published

#	Model	Score
1	Muse SparkMeta	75.5%
2	Muse Spark 1.1Meta	75.3%
3	Gemini 3.1 Pro Previewas “gemini-3.1-pro-preview”	71.4%
4	GPT-5.4 Proas “gpt-5.4-pro-2026-03-05”	69.2%
5	Gemini 3 Proas “gemini-3-pro-preview”	65.7%
6	GPT-5.1as “gpt-5.1-2025-11-13-thinking”	63.4%
7	GPT-5as “gpt-5-thinking”	63.2%
8	o3-proas “o3-pro-2025-06-10-reasoning-high”	62.4%
9	Kimi K2.5as “kimi-k2.5”	61.4%
10	Gemini 3.1 Flash-Liteas “gemini-3.1-flash-lite-preview”	60.6%
11	GPT-5 minias “gpt-5-mini-thinking”	59.0%
12	Claude Opus 4.5as “claude-opus-4-5-20251101-thinking”	59.0%
13	Claude Opus 4as “claude-4-opus-thinking”	58.6%
14	Claude Opus 4.1as “claude-opus-4-1-20250805-thinking”	57.2%
15	Claude Sonnet 4as “claude-4-sonnet-thinking”	57.1%
16	o3as “o3-2025-04-16-reasoning-high”	56.6%
17	Claude Opus 4.6as “claude-opus-4-6 (Non-Thinking)”	56.0%
18	Kimi K2 Thinkingas “kimi-k2-thinking”	55.4%
19	Claude Sonnet 4.5as “claude-sonnet-4-5-20250929-thinking”	55.3%
20	Gemini 2.5 Pro (Jun 2025)as “gemini-2-5-pro”	53.6%
21	Claude 3 7 Sonnetas “claude-3-7-sonnet-thinking”	51.6%
22	GPT-5.1 Instantas “gpt-5.1-2025-11-13-instant”	51.2%
23	Claude 4.5 Haikuas “claude-haiku-4-5-20251001-thinking”	50.5%
24	DeepSeek V3P1as “deepseek-v3p1”	46.1%
25	gpt-oss-120bOpenAI	45.3%

About this board and its history

Multi-turn conversations that test instruction retention, inference memory and self-coherence.

Scores as published by Scale AI on the capture date (2026-09-24). Source: scale.com leaderboard page.