gpt-sw3-126m [float16] [rev: c9d5a2f3fe90]: benchmark results
AI Sweden's gpt-sw3-126m, a decoder-only causal language model developed with RISE and WASP WARA for Media and Language using NeMo Megatron GPT. Its card describes Nordic-language, English and programming-code pretraining. This row preserves revision c9d5a2f3fe905557cf0acba496a903255a11907c and its explicit float16 evaluation setting; matching family names do not establish checkpoint equivalence. Provider: AI Sweden. Access: Open.
Aggregate Rating 1083 ± 43, rank #10775 of 10870 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 44.07 | MC2 (%) (0-shot) | 25.9 |
| Open LLM Leaderboard v1 - GSM8K | 0.08 | Accuracy (%) (5-shot) | 9.6 |
| Open LLM Leaderboard v1 - HellaSwag | 29.56 | Normalized accuracy (%) (10-shot) | 5.3 |
| Open LLM Leaderboard v1 - MMLU | 24.53 | Accuracy (%) (5-shot) | 4.1 |
| Open LLM Leaderboard v1 - WinoGrande | 50.43 | Accuracy (%) (5-shot) | 3.6 |
| Open LLM Leaderboard v1 - ARC Challenge | 22.01 | Normalized accuracy (%) (25-shot) | 0.8 |
Interactive version: theaggregate.ai/model?slug=gpt-sw3-126m-float16-rev-c9d5a2f3fe90 · How It Works · Data refreshed daily, snapshot 2026-11-12.