Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2510.26550 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908643392225280 |
|---|---|
| author | FitzGerald, Jack Lazaridis, Aristotelis Bates, Dylan Sharma, Aman Castillo, Jonnathan Azami, Yousif Bailey, Sean Cao, Jeremy Damianov, Peter de Haan, Kevin Kerbs, Luke Lu, Vincent Madigan, Joseph McLaurin, Jeremy Tainer, Jonathan Anderson, Dave Beck, Jonathan Cuticello, Jamie Malkerson, Colton Saltsman, Tyler |
| author_facet | FitzGerald, Jack Lazaridis, Aristotelis Bates, Dylan Sharma, Aman Castillo, Jonnathan Azami, Yousif Bailey, Sean Cao, Jeremy Damianov, Peter de Haan, Kevin Kerbs, Luke Lu, Vincent Madigan, Joseph McLaurin, Jeremy Tainer, Jonathan Anderson, Dave Beck, Jonathan Cuticello, Jamie Malkerson, Colton Saltsman, Tyler |
| contents | We present EdgeRunner 20B, a fine-tuned version of gpt-oss-20b optimized for military tasks. EdgeRunner 20B was trained on 1.6M high-quality records curated from military documentation and websites. We also present four new tests sets: (a) combat arms, (b) combat medic, (c) cyber operations, and (d) mil-bench-5k (general military knowledge). On these military test sets, EdgeRunner 20B matches or exceeds GPT-5 task performance with 95%+ statistical significance, except for the high reasoning setting on the combat medic test set and the low reasoning setting on the mil-bench-5k test set. Versus gpt-oss-20b, there is no statistically-significant regression on general-purpose benchmarks like ARC-C, GPQA Diamond, GSM8k, IFEval, MMLU Pro, or TruthfulQA, except for GSM8k in the low reasoning setting. We also present analyses on hyperparameter settings, cost, and throughput. These findings show that small, locally-hosted models are ideal solutions for data-sensitive operations such as in the military domain, allowing for deployment in air-gapped edge devices. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_26550 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | EdgeRunner 20B: Military Task Parity with GPT-5 while Running on the Edge FitzGerald, Jack Lazaridis, Aristotelis Bates, Dylan Sharma, Aman Castillo, Jonnathan Azami, Yousif Bailey, Sean Cao, Jeremy Damianov, Peter de Haan, Kevin Kerbs, Luke Lu, Vincent Madigan, Joseph McLaurin, Jeremy Tainer, Jonathan Anderson, Dave Beck, Jonathan Cuticello, Jamie Malkerson, Colton Saltsman, Tyler Artificial Intelligence We present EdgeRunner 20B, a fine-tuned version of gpt-oss-20b optimized for military tasks. EdgeRunner 20B was trained on 1.6M high-quality records curated from military documentation and websites. We also present four new tests sets: (a) combat arms, (b) combat medic, (c) cyber operations, and (d) mil-bench-5k (general military knowledge). On these military test sets, EdgeRunner 20B matches or exceeds GPT-5 task performance with 95%+ statistical significance, except for the high reasoning setting on the combat medic test set and the low reasoning setting on the mil-bench-5k test set. Versus gpt-oss-20b, there is no statistically-significant regression on general-purpose benchmarks like ARC-C, GPQA Diamond, GSM8k, IFEval, MMLU Pro, or TruthfulQA, except for GSM8k in the low reasoning setting. We also present analyses on hyperparameter settings, cost, and throughput. These findings show that small, locally-hosted models are ideal solutions for data-sensitive operations such as in the military domain, allowing for deployment in air-gapped edge devices. |
| title | EdgeRunner 20B: Military Task Parity with GPT-5 while Running on the Edge |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2510.26550 |