SEAL - MultiChallenge: leaderboard
Multi-turn conversational capabilities across four dimensions: instruction retention, context recall, iterative document editing, and self-consistency in extended dialogues.
Metric: Score. Source: scale.com. Saturation forecast: Around January 2027. 32 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6.1 Sol | 81.95 | #7 |
| 2 | Muse Spark | 75.52 | #64 |
| 3 | Muse Spark 1.1 | 75.3 | #52 |
| 4 | Gemini 3.1 Pro (Preview) | 71.37 | #58 |
| 5 | Claude Haiku 5.5 | 67.67 | #40 |
| 6 | Gemini 3 Pro (Preview) | 65.67 | #66 |
| 7 | GPT-5 (Thinking) | 63.19 | #98 (GPT-5) |
| 8 | Kimi K2.5 | 61.39 | #144 |
| 9 | Gemini 3.1 Flash Lite (Preview) | 60.61 | #194 |
| 10 | Claude Opus 4.5 (20251101) (Thinking) | 58.97 | #74 (Claude Opus 4.5 (20251101)) |
| 11 | Claude Opus 4 (Thinking) | 58.62 | #163 (Claude Opus 4) |
| 12 | Claude Opus 4.1 (20250805) (Thinking) | 57.2 | #132 (Claude Opus 4.1 (20250805)) |
| 13 | Claude Sonnet 4 (Thinking) | 57.11 | #199 (Claude Sonnet 4) |
| 14 | O3 (2025-04-16) (High) | 56.62 | #121 (O3 (2025-04-16)) |
| 15 | Claude Opus 4.6 (Non-reasoning) | 56.02 | #63 (Claude Opus 4.6) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #8 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=seal-multichallenge · How It Works · Data refreshed daily, snapshot 2026-10-20.