Ko-WideSearch - Row F1: leaderboard
Evaluates Korean broad web research with structured table outputs on 228 tables in a single-pass protocol. Item, column, row and complete-table success are distinct published metrics; parse rate is not an extra board. This series reports Row F1 using F1 (%). Source: the original paper result table.
Metric: F1 (%). Source: arxiv.org. Saturation forecast: Around February 2027. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 53.7 | #33 |
| 2 | Claude Opus 4.8 | 52.9 | #39 |
| 3 | Claude Opus 4.7 | 51.6 | #51 |
| 4 | Claude Opus 4.6 | 48.9 | #67 |
| 5 | Gemini 3.1 Pro (Preview) | 45.9 | #64 |
| 6 | Gemini 3.1 Flash Lite | 45.9 | #200 |
| 7 | DeepSeek V4 Pro | 45 | #112 |
| 8 | Gemini 3.5 Flash | 44.1 | #66 |
| 9 | Claude Sonnet 4.6 | 43.6 | #94 |
| 10 | GPT-5.4 | 41.6 | #83 |
| 11 | GLM-5.1 | 34 | #109 |
| 12 | GPT-5.4 Mini | 33.3 | #228 |
| 13 | Claude Haiku 4.5 | 28.8 | #287 |
| 14 | A.X 4.0 | 24.2 | #436 |
| 15 | Gemma 4 31B | 23 | #212 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ko-widesearch-row-f1 · How It Works · Data refreshed daily, snapshot 2026-11-05.