K2S3-Mistral-7B-v1.43: benchmark results
K2S3 release v1.43 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 1407 ± 33, rank #5755 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - GSM8K | 54.28 | Accuracy (%) (5-shot) | 66 |
| Open LLM Leaderboard v1 - WinoGrande | 78.77 | Accuracy (%) (5-shot) | 65.4 |
| Open LLM Leaderboard v1 - MMLU | 62.96 | Accuracy (%) (5-shot) | 57.7 |
| Open LLM Leaderboard v1 - HellaSwag | 83.22 | Normalized accuracy (%) (10-shot) | 56.9 |
| Open LLM Leaderboard v1 - ARC Challenge | 61.26 | Normalized accuracy (%) (25-shot) | 52 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 50.49 | MC2 (%) (0-shot) | 48.5 |
Interactive version: theaggregate.ai/model?slug=k2s3-mistral-7b-v1-43 · How It Works · Data refreshed daily, snapshot 2026-11-08.