BTZSC - Topic: leaderboard
BTZSC topic slice measures macro-F1 on zero-shot topic classification datasets.
Metric: Macro-F1 (%). Source: huggingface.co. Saturation forecast: Around January 2027. 35 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral Nemo Instruct (2407) | 69.05 | #936 |
| 2 | Qwen 3 8B | 64.94 | #674 |
| 3 | Qwen 3 4B | 63.83 | #829 |
| 4 | Qwen3-Embedding-0.6B | 48.56 | #802 |
| 5 | Qwen3-Embedding-8B | 44.35 | #685 |
| 6 | Phi-4 Mini Instruct | 43.91 | #1245 |
| 7 | Llama 3.2 3B Instruct | 43.71 | #1352 |
| 8 | Gemma 3 1B (IT) | 34.46 | #1504 |
| 9 | Gemma 3 270M (IT) | 28.73 | #1581 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=btzsc-topic · How It Works · Data refreshed daily, snapshot 2026-10-20.