Ko-WideSearch - Table Success: 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 Table Success using Table Success (%). Source: the original paper result table.
Metric: Table Success (%). Source: arxiv.org. Saturation forecast: Around July 2028. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 19.3 | #33 |
| 2 | Claude Opus 4.8 | 16.2 | #39 |
| 3 | Claude Opus 4.7 | 15.8 | #51 |
| 4 | Claude Opus 4.6 | 14.9 | #67 |
| 5 | Gemini 3.1 Pro (Preview) | 14.5 | #64 |
| 6 | Gemini 3.5 Flash | 12.7 | #66 |
| 7 | DeepSeek V4 Pro | 12.3 | #112 |
| 8 | GLM-5.1 | 12.3 | #109 |
| 9 | Claude Sonnet 4.6 | 11.8 | #94 |
| 10 | GPT-5.4 | 10.5 | #83 |
| 11 | Claude Haiku 4.5 | 7 | #287 |
| 12 | Gemini 3.1 Flash Lite | 7 | #200 |
| 13 | GPT-5.4 Mini | 5.7 | #228 |
| 14 | GPT-5.4 Nano | 5.7 | #359 |
| 15 | A.X 4.0 | 4.4 | #436 |
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-table-success · How It Works · Data refreshed daily, snapshot 2026-11-05.