Yandex AliceAI Cards - GSM8K (CoT): leaderboard
GSM8K grade-school math word problems, 8-shot with chain of thought, run by Yandex. Yandex AliceAI-T5-35B-A0.6B model card (September 2026); pretrained base checkpoints; accuracy in percent, higher is better.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 35B A3B Base | 90.4 | #671 |
| 2 | Qwen3.5-4B-Base | 84.5 | #2368 |
| 3 | Qwen3.5-2B-Base | 69.5 | #6809 |
| 4 | Gemma 4 E4B | 58.4 | #2217 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=yandex-aliceai-cards-gsm8k-cot · How It Works · Data refreshed daily, snapshot 2026-11-12.