K2S3-Mistral-7B-v1.46: benchmark results
K2S3 release v1.46 published by Changgil: a full-parameter supervised fine-tune of mistralai/Mistral-7B-v0.1, using alpaca-gpt4-data and OpenOrca samples. The release configuration uses 32 Mistral decoder layers; versioned siblings are not interchangeable checkpoints. Provider: Changgil. Access: Open.
Aggregate Rating 1405 ± 33, rank #5824 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - WinoGrande | 80.43 | Accuracy (%) (5-shot) | 73.8 |
| Open LLM Leaderboard v1 - HellaSwag | 83.49 | Normalized accuracy (%) (10-shot) | 59.1 |
| Open LLM Leaderboard v1 - GSM8K | 42.38 | Accuracy (%) (5-shot) | 57 |
| Open LLM Leaderboard v1 - MMLU | 62.81 | Accuracy (%) (5-shot) | 56.8 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 51.82 | MC2 (%) (0-shot) | 53.4 |
| Open LLM Leaderboard v1 - ARC Challenge | 60.75 | Normalized accuracy (%) (25-shot) | 49.3 |
Interactive version: theaggregate.ai/model?slug=k2s3-mistral-7b-v1-46 · How It Works · Data refreshed daily, snapshot 2026-11-08.