SEA-HELM (Thai) - SEA-MT-Bench (LLM Judge): leaderboard
SEA-MT-Bench: multi-turn chat questions written in Thai, with responses graded by an LLM judge. Normalized 0-100 score from AI Singapore's SEA-HELM leaderboard: (score - random baseline) / (maximum - baseline), floored at 0 and averaged over eight runs; higher is better.
Metric: Normalized Score. Source: leaderboard.sea-lion.ai. Saturation forecast: Estimated already saturated. 58 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 122B A10B (Reasoning) | 85.07 | #184 (Qwen 3.5 122B A10B) |
| 2 | Gemma 4 26B A4B (IT) (Reasoning) | 83.99 | #293 (Gemma 4 26B A4B (IT)) |
| 3 | Qwen 3.5 35B A3B (Reasoning) | 83.78 | #268 (Qwen 3.5 35B A3B) |
| 4 | Qwen 3.5 27B (Reasoning) | 83.77 | #192 (Qwen 3.5 27B) |
| 5 | GPT-OSS-120B (Reasoning) | 83.21 | #346 (GPT-OSS-120B) |
| 6 | Qwen 3.6 27B (Reasoning) | 82.71 | #182 (Qwen 3.6 27B) |
| 7 | Qwen 3.6 35B A3B (Reasoning) | 82.37 | #238 (Qwen 3.6 35B A3B) |
| 8 | Qwen 3 VL 32B Instruct | 81.43 | #290 |
| 9 | Qwen 3 32B (Thinking) | 81.29 | #449 (Qwen 3 32B) |
| 10 | Qwen 3.5 9B (Reasoning) | 81.03 | #391 (Qwen 3.5 9B) |
| 11 | Qwen 3 14B (Reasoning) | 79.35 | #542 (Qwen 3 14B) |
| 12 | Qwen 3 VL 8B Instruct | 78.28 | #419 |
| 13 | Qwen 3.5 4B (Reasoning) | 76.11 | #514 (Qwen 3.5 4B) |
| 14 | Qwen 3 8B (Thinking) | 76.08 | #678 (Qwen 3 8B) |
| 15 | Gemma 3 27B (IT) | 75.45 | #528 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=sea-helm-thai-sea-mt-bench-llm-judge · How It Works · Data refreshed daily, snapshot 2026-11-05.