Horangi 4 - BigCodeBench (100-task subset): leaderboard
A 100-task subset of BigCodeBench library-heavy Python programming tasks, as run by the Horangi Korean LLM Leaderboard 4 (W&B Korea). Pass rate (%) against the tests, higher is better. Older runs have this score in their summaries only.
Metric: Pass rate (%). Source: horangi.ai. Saturation forecast: Around July 2028. 105 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5 (High) | 68 | #20 (Claude Fable 5) |
| 2 | GPT-5.6 Sol (Max) | 63 | #22 (GPT-5.6 Sol) |
| 3 | Claude Opus 4.8 (High) | 62 | #39 (Claude Opus 4.8) |
| 4 | GPT-5.6 Terra (Max) | 60 | #40 (GPT-5.6 Terra) |
| 5 | GPT-5.5 (xHigh) | 59 | #33 (GPT-5.5) |
| 6 | Claude Sonnet 5 (High) | 59 | #65 (Claude Sonnet 5) |
| 7 | GPT-5.4 Mini (2026-03-17) (xHigh) | 59 | #180 (GPT-5.4 Mini (2026-03-17)) |
| 8 | GPT-5.4 (2026-03-05) (xHigh) | 58 | #63 (GPT-5.4 (2026-03-05)) |
| 9 | Qwen 3.7 Plus | 57.14 | #98 |
| 10 | Gemini 3.5 Flash | 57 | #66 |
| 11 | Claude Opus 4.6 (High) | 57 | #67 (Claude Opus 4.6) |
| 12 | GPT-5.2 (2025-12-11) (xHigh) | 56.57 | |
| 13 | GPT-5.4 Nano (2026-03-17) (xHigh) | 56.57 | #285 (GPT-5.4 Nano (2026-03-17)) |
| 14 | GPT-5.1 (2025-11-13) | 56 | #107 |
| 15 | Claude Opus 4.1 (20250805) | 56 | #139 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=horangi-4-bigcodebench-100-task-subset · How It Works · Data refreshed daily, snapshot 2026-11-05.