1,408
Overall · rank 133 of 177
- Unit
- Arena rating, higher is better
- Range
- 1,404 to 1,411
- Sample
- 30215 votes
- Configuration
- Qwen3.5-35B-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.
1,403
Business, management and finance · rank 119 of 402
- Unit
- Arena rating, higher is better
- Range
- 1,395 to 1,411
- Sample
- 6114 votes
- Configuration
- Qwen3.5-35B-A3B
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,342
Creative writing · rank 159 of 407
- Unit
- Arena rating, higher is better
- Range
- 1,333 to 1,352
- Sample
- 4668 votes
- Configuration
- Qwen3.5-35B-A3B
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,420
Expert prompts · rank 115 of 359
- Unit
- Arena rating, higher is better
- Range
- 1,408 to 1,431
- Sample
- 2787 votes
- Configuration
- Qwen3.5-35B-A3B
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,386
Instruction following · rank 141 of 409
- Unit
- Arena rating, higher is better
- Range
- 1,380 to 1,393
- Sample
- 9917 votes
- Configuration
- Qwen3.5-35B-A3B
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,394
Overall · rank 151 of 409
- Unit
- Arena rating, higher is better
- Range
- 1,390 to 1,398
- Sample
- 30420 votes
- Configuration
- Qwen3.5-35B-A3B
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,364
Writing, literature and language · rank 154 of 408
- Unit
- Arena rating, higher is better
- Range
- 1,356 to 1,371
- Sample
- 7034 votes
- Configuration
- Qwen3.5-35B-A3B
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
18.0%
Requested-length error · rank 150 of 370
- Unit
- % off the requested word count, lower is better
- Configuration
- Qwen3.5-35B-A3B (think-prefill reasoning)
- Measured
- 14 Mar 2026
- Not shown
- Not other format limits such as character counts or bullet counts.
17.0%
Requested-length error · rank 140 of 370
- Unit
- % off the requested word count, lower is better
- Configuration
- Qwen3.5-35B-A3B (no reasoning)
- Measured
- 14 Mar 2026
- Not shown
- Not other format limits such as character counts or bullet counts.
0.28
Style adherence · rank 357 of 370
- Unit
- score from 0 to 1, higher is better
- Configuration
- Qwen3.5-35B-A3B (think-prefill reasoning)
- Measured
- 14 Mar 2026
- Not shown
- Not brand voice on your own examples: UGI's prompts are private and lean towards creative writing.
0.33
Style adherence · rank 254 of 370
- Unit
- score from 0 to 1, higher is better
- Configuration
- Qwen3.5-35B-A3B (no reasoning)
- Measured
- 14 Mar 2026
- Not shown
- Not brand voice on your own examples: UGI's prompts are private and lean towards creative writing.
42.6
Writing score · rank 193 of 370
- Unit
- score out of 100, higher is better
- Configuration
- Qwen3.5-35B-A3B (think-prefill reasoning)
- Measured
- 14 Mar 2026
- Not shown
- Not business copy quality: UGI's prompts are private and lean towards creative writing, and models that often refuse get no score.
37.0
Writing score · rank 231 of 370
- Unit
- score out of 100, higher is better
- Configuration
- Qwen3.5-35B-A3B (no reasoning)
- Measured
- 14 Mar 2026
- Not shown
- Not business copy quality: UGI's prompts are private and lean towards creative writing, and models that often refuse get no score.
99.8%
Answer rate · rank 34 of 108
- Unit
- % of documents, higher is better
- Configuration
- Qwen3.5-35B-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.
10.5%
Hallucination rate · rank 62 of 108
- Unit
- % of summaries, lower is better
- Configuration
- Qwen3.5-35B-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.
