1,362
Overall · rank 161 of 177
- Unit
- Arena rating, higher is better
- Range
- 1,357 to 1,367
- Sample
- 15422 votes
- Configuration
- Nemotron 3 Nano 30B A3B
- Measured
- 25 Sep 2026
- Not shown
- Not an error rate: claims that cannot be checked on the web are skipped, and preference still carries most of the weight.
1,317
Business, management and finance · rank 218 of 402
- Unit
- Arena rating, higher is better
- Range
- 1,306 to 1,328
- Sample
- 3004 votes
- Configuration
- Nemotron 3 Nano 30B A3B
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,241
Creative writing · rank 271 of 407
- Unit
- Arena rating, higher is better
- Range
- 1,229 to 1,254
- Sample
- 2412 votes
- Configuration
- Nemotron 3 Nano 30B A3B
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,331
Expert prompts · rank 201 of 359
- Unit
- Arena rating, higher is better
- Range
- 1,312 to 1,350
- Sample
- 1041 votes
- Configuration
- Nemotron 3 Nano 30B A3B
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,291
Instruction following · rank 248 of 409
- Unit
- Arena rating, higher is better
- Range
- 1,282 to 1,300
- Sample
- 4365 votes
- Configuration
- Nemotron 3 Nano 30B A3B
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,313
Overall · rank 247 of 409
- Unit
- Arena rating, higher is better
- Range
- 1,308 to 1,319
- Sample
- 15731 votes
- Configuration
- Nemotron 3 Nano 30B A3B
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,267
Writing, literature and language · rank 267 of 408
- Unit
- Arena rating, higher is better
- Range
- 1,256 to 1,277
- Sample
- 3585 votes
- Configuration
- Nemotron 3 Nano 30B A3B
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
15.5%
Average · rank 34 of 34
- Unit
- % resolved, higher is better
- Configuration
- Nemotron 3 Nano 30B A3B (openhands)
- Measured
- 30 Jun 2026
- Not shown
- Not the model alone: the score is for the model inside the OpenHands agent, at the SDK version shown.
20.9%
Front end (SWE-Bench Multimodal) · rank 32 of 34
- Unit
- % resolved, higher is better
- Configuration
- Nemotron 3 Nano 30B A3B (openhands)
- Measured
- 30 Jun 2026
- Not shown
- Not the model alone: the score is for the model inside the OpenHands agent, at the SDK version shown.
6.2%
Greenfield (Commit0) · rank 33 of 34
- Unit
- % resolved, higher is better
- Configuration
- Nemotron 3 Nano 30B A3B (openhands)
- Measured
- 30 Jun 2026
- Not shown
- Not the model alone: the score is for the model inside the OpenHands agent, at the SDK version shown.
8.5%
Information gathering (GAIA) · rank 34 of 34
- Unit
- % resolved, higher is better
- Configuration
- Nemotron 3 Nano 30B A3B (openhands)
- Measured
- 30 Jun 2026
- Not shown
- Not the model alone: the score is for the model inside the OpenHands agent, at the SDK version shown.
34.2%
Issue resolution (SWE-Bench) · rank 34 of 34
- Unit
- % resolved, higher is better
- Configuration
- Nemotron 3 Nano 30B A3B (openhands)
- Measured
- 30 Jun 2026
- Not shown
- Not the model alone: the score is for the model inside the OpenHands agent, at the SDK version shown.
7.6%
Testing (SWT-Bench) · rank 34 of 34
- Unit
- % resolved, higher is better
- Configuration
- Nemotron 3 Nano 30B A3B (openhands)
- Measured
- 30 Jun 2026
- Not shown
- Not the model alone: the score is for the model inside the OpenHands agent, at the SDK version shown.
35.0%
Requested-length error · rank 277 of 370
- Unit
- % off the requested word count, lower is better
- Configuration
- Nemotron 3 Nano 30B A3B
- Measured
- 16 Dec 2025
- Not shown
- Not other format limits such as character counts or bullet counts.
251.0%
Requested-length error · rank 369 of 370
- Unit
- % off the requested word count, lower is better
- Configuration
- Nemotron 3 Nano 30B A3B (think-prefill reasoning)
- Measured
- 18 Dec 2025
- Not shown
- Not other format limits such as character counts or bullet counts.
0.28
Style adherence · rank 355 of 370
- Unit
- score from 0 to 1, higher is better
- Configuration
- Nemotron 3 Nano 30B A3B
- Measured
- 16 Dec 2025
- Not shown
- Not brand voice on your own examples: UGI's prompts are private and lean towards creative writing.
0.24
Style adherence · rank 370 of 370
- Unit
- score from 0 to 1, higher is better
- Configuration
- Nemotron 3 Nano 30B A3B (think-prefill reasoning)
- Measured
- 18 Dec 2025
- Not shown
- Not brand voice on your own examples: UGI's prompts are private and lean towards creative writing.
25.9
Writing score · rank 297 of 370
- Unit
- score out of 100, higher is better
- Configuration
- Nemotron 3 Nano 30B A3B
- Measured
- 16 Dec 2025
- Not shown
- Not business copy quality: UGI's prompts are private and lean towards creative writing, and models that often refuse get no score.
3.4
Writing score · rank 370 of 370
- Unit
- score out of 100, higher is better
- Configuration
- Nemotron 3 Nano 30B A3B (think-prefill reasoning)
- Measured
- 18 Dec 2025
- Not shown
- Not business copy quality: UGI's prompts are private and lean towards creative writing, and models that often refuse get no score.
99.6%
Answer rate · rank 48 of 108
- Unit
- % of documents, higher is better
- Configuration
- Nemotron 3 Nano 30B A3B
- Measured
- 22 Sep 2026
- Not shown
- Not a quality score: a low rate usually means content filters were triggered, and hallucination rates are measured on answered documents only.
9.6%
Hallucination rate · rank 55 of 108
- Unit
- % of summaries, lower is better
- Configuration
- Nemotron 3 Nano 30B A3B
- Measured
- 22 Sep 2026
- Not shown
- Not errors in open questions or other tasks: only summarisation, judged by Vectara's own model (HHEM-2.3), not by people, on news-style documents rather than your data.
