MultiChallenge-LILT - Korean: leaderboard
MultiChallenge-LILT - Korean: Pass@1 accuracy (%) in LILT Aurora's localized evaluation. Multi-turn conversational instruction following; target model effort unreported.
Metric: Pass@1 accuracy (%). Source: aurora.lilt.com. Saturation forecast: Rough model projection: around 2026. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.8 Flash (Aurora Run: MultiChallenge-LILT; effort unreported) | 60.2 | |
| 2 | Claude Opus 5.5 (Aurora Run: MultiChallenge-LILT; effort unreported) | 57.85 | |
| 3 | GPT-5.6 Sol (Aurora Run: MultiChallenge-LILT; effort unreported) | 54.58 | |
| 4 | Claude Opus 5 (Aurora Run: MultiChallenge-LILT; effort unreported) | 54.13 | |
| 5 | GPT-6 Sol (Aurora Run: MultiChallenge-LILT; effort unreported) | 51.44 | |
| 6 | Nemotron 3 Ultra (Aurora Run: MultiChallenge-LILT; effort unreported) | 50.29 | |
| 7 | Kimi K3 (Aurora Run: MultiChallenge-LILT; effort unreported) | 43.11 | |
| 8 | Command A+ (Aurora Run: MultiChallenge-LILT; effort unreported) | 40.86 | |
| 9 | Mistral Medium 3.5 (Aurora Run: MultiChallenge-LILT; effort unreported) | 37.64 |
Interactive version: theaggregate.ai/benchmark?slug=multichallenge-lilt-korean · How It Works · Data refreshed daily, snapshot 2026-11-05.