LKValues: leaderboard
Micro-averaged A/B/BOTH/0 accuracy across evaluated English and Sinhala LKvaluesBench conditions and Sri Lankan-specific and Universal prompt framings. Scores are fractions from 0 to 1; base and LKValues-adapted checkpoints remain distinct.
Metric: Micro-averaged Accuracy (0-1, English/Sinhala, both prompts). Source: arxiv.org. Saturation forecast: Estimated already saturated. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 4 31B (IT) | 0.95 | #176 |
| 2 | Kimi K2 0905 | 0.93 | #288 |
| 3 | Grok 4.3 | 0.92 | #188 |
| 4 | DeepSeek V3 | 0.91 | #321 |
| 5 | Gemma 3 27B (IT) | 0.85 | #523 |
| 6 | Qwen 3 235B A22B | 0.75 | #313 |
| 7 | Qwen 3 8B | 0.75 | #674 |
| 8 | Llama 3.3 70B Instruct | 0.73 | #534 |
| 9 | Qwen3.5-4B-Base | 0.61 | #798 |
| 10 | c4ai-command-r-08-2024 | 0.5 | #921 |
| 11 | Llama 3.1 8B Instruct | 0.35 | #1035 |
| 12 | Tiny Aya Global | 0.29 | #1325 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=lkvalues · How It Works · Data refreshed daily, snapshot 2026-10-20.