1,425
Overall · rank 36 of 44
- Unit
- Arena rating, higher is better
- Range
- 1,419 to 1,431
- Sample
- 28212 votes
- Configuration
- GPT-5.2
- Measured
- 13 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,423
Overall · rank 35 of 44
- Unit
- Arena rating, higher is better
- Range
- 1,414 to 1,433
- Sample
- 7108 votes
- Configuration
- GPT-5.2 (high)
- Measured
- 13 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,162
Overall · rank 24 of 34
- Unit
- Arena rating, higher is better
- Range
- 1,156 to 1,167
- Sample
- 52712 votes
- Configuration
- GPT-5.2
- Measured
- 24 Aug 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,137
Overall · rank 28 of 34
- Unit
- Arena rating, higher is better
- Range
- 1,131 to 1,143
- Sample
- 75658 votes
- Configuration
- GPT-5.2
- Measured
- 24 Aug 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,442
Overall · rank 63 of 177
- Unit
- Arena rating, higher is better
- Range
- 1,440 to 1,445
- Sample
- 82944 votes
- Configuration
- GPT-5.2
- 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.
1,442
Overall · rank 65 of 177
- Unit
- Arena rating, higher is better
- Range
- 1,438 to 1,445
- Sample
- 49030 votes
- Configuration
- GPT-5.2 (high)
- 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.
1,444
Business, management and finance · rank 60 of 402
- Unit
- Arena rating, higher is better
- Range
- 1,438 to 1,450
- Sample
- 16851 votes
- Configuration
- GPT-5.2
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,443
Business, management and finance · rank 60 of 402
- Unit
- Arena rating, higher is better
- Range
- 1,437 to 1,450
- Sample
- 9768 votes
- Configuration
- GPT-5.2 (high)
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,391
Creative writing · rank 94 of 407
- Unit
- Arena rating, higher is better
- Range
- 1,385 to 1,397
- Sample
- 13750 votes
- Configuration
- GPT-5.2
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,387
Creative writing · rank 97 of 407
- Unit
- Arena rating, higher is better
- Range
- 1,379 to 1,394
- Sample
- 7645 votes
- Configuration
- GPT-5.2 (high)
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,459
Expert prompts · rank 73 of 359
- Unit
- Arena rating, higher is better
- Range
- 1,452 to 1,467
- Sample
- 7976 votes
- Configuration
- GPT-5.2
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,475
Expert prompts · rank 41 of 359
- Unit
- Arena rating, higher is better
- Range
- 1,465 to 1,485
- Sample
- 3993 votes
- Configuration
- GPT-5.2 (high)
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,424
Instruction following · rank 83 of 409
- Unit
- Arena rating, higher is better
- Range
- 1,419 to 1,429
- Sample
- 27771 votes
- Configuration
- GPT-5.2
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,427
Instruction following · rank 76 of 409
- Unit
- Arena rating, higher is better
- Range
- 1,421 to 1,433
- Sample
- 15266 votes
- Configuration
- GPT-5.2 (high)
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,436
Overall · rank 86 of 409
- Unit
- Arena rating, higher is better
- Range
- 1,432 to 1,439
- Sample
- 83555 votes
- Configuration
- GPT-5.2
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,437
Overall · rank 83 of 409
- Unit
- Arena rating, higher is better
- Range
- 1,434 to 1,441
- Sample
- 49570 votes
- Configuration
- GPT-5.2 (high)
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,411
Writing, literature and language · rank 84 of 408
- Unit
- Arena rating, higher is better
- Range
- 1,406 to 1,417
- Sample
- 20281 votes
- Configuration
- GPT-5.2
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
1,408
Writing, literature and language · rank 88 of 408
- Unit
- Arena rating, higher is better
- Range
- 1,402 to 1,414
- Sample
- 11497 votes
- Configuration
- GPT-5.2 (high)
- Measured
- 25 Sep 2026
- Not shown
- Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
79.4%
Irrelevance detection · rank 61 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not general refusal or safety behaviour.
87.3%
Irrelevance detection · rank 21 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not general refusal or safety behaviour.
45.8%
Memory · rank 13 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not long-term personal memory in a product: sessions are BFCL's scripted ones.
3.9%
Memory · rank 92 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not long-term personal memory in a product: sessions are BFCL's scripted ones.
28.1%
Multi-turn tasks · rank 42 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not open-ended agent work: the tools and tasks are BFCL's simulated APIs.
43.8%
Multi-turn tasks · rank 24 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not open-ended agent work: the tools and tasks are BFCL's simulated APIs.
55.9%
Overall accuracy · rank 16 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not a neutral average: the weighting is BFCL's. Not reliability on your own tools: BFCL's functions and queries are a fixed test set, and a correct call is judged by its form, not by what it achieved.
45.3%
Overall accuracy · rank 38 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not a neutral average: the weighting is BFCL's. Not reliability on your own tools: BFCL's functions and queries are a fixed test set, and a correct call is judged by its form, not by what it achieved.
75.0%
Relevance detection · rank 62 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not whether the call itself was right; only that one was attempted.
75.0%
Relevance detection · rank 62 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not whether the call itself was right; only that one was attempted.
81.8%
Single-turn calls (curated) · rank 67 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not reliability on your own tools: BFCL's functions and queries are a fixed test set, and a correct call is judged by its form, not by what it achieved.
78.3%
Single-turn calls (curated) · rank 79 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not reliability on your own tools: BFCL's functions and queries are a fixed test set, and a correct call is judged by its form, not by what it achieved.
70.4%
Single-turn calls (user-contributed) · rank 58 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not reliability on your own tools: BFCL's functions and queries are a fixed test set, and a correct call is judged by its form, not by what it achieved.
67.1%
Single-turn calls (user-contributed) · rank 73 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not reliability on your own tools: BFCL's functions and queries are a fixed test set, and a correct call is judged by its form, not by what it achieved.
75.5%
Web search · rank 10 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not general research quality: questions have short, checkable answers.
40.5%
Web search · rank 30 of 109
- Unit
- % correct, higher is better
- Configuration
- gpt-5.2
- Measured
- 16 Dec 2025
- Not shown
- Not general research quality: questions have short, checkable answers.
58.8%
Average · rank 10 of 34
- Unit
- % resolved, higher is better
- Configuration
- gpt-5.2
- 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.
30.9%
Front end (SWE-Bench Multimodal) · rank 20 of 34
- Unit
- % resolved, higher is better
- Configuration
- gpt-5.2
- 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.
50.0%
Greenfield (Commit0) · rank 6 of 34
- Unit
- % resolved, higher is better
- Configuration
- gpt-5.2
- 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.
65.5%
Information gathering (GAIA) · rank 16 of 34
- Unit
- % resolved, higher is better
- Configuration
- gpt-5.2
- 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.
74.6%
Issue resolution (SWE-Bench) · rank 13 of 34
- Unit
- % resolved, higher is better
- Configuration
- gpt-5.2
- 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.
73.2%
Testing (SWT-Bench) · rank 10 of 34
- Unit
- % resolved, higher is better
- Configuration
- gpt-5.2
- 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.
72.0%
Consistency (pass^4) · rank 1 of 8
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 5 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
22.0%
Consistency (pass^4) · rank 8 of 8
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 26 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
83.0%
Task success (pass^1) · rank 2 of 8
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 5 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
52.5%
Task success (pass^1) · rank 8 of 8
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 26 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
18.6%
Consistency (pass^4) · rank 11 of 21
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 5 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
32.2%
Task success (pass^1) · rank 13 of 21
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 5 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
4.1%
Consistency (pass^4) · rank 1 of 1
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 26 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
12.6%
Task success (pass^1) · rank 1 of 1
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 26 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
51.8%
Consistency (pass^4) · rank 2 of 8
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 5 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
45.6%
Consistency (pass^4) · rank 6 of 8
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 26 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
81.6%
Task success (pass^1) · rank 2 of 8
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 5 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
75.0%
Task success (pass^1) · rank 6 of 8
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 26 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
71.9%
Consistency (pass^4) · rank 4 of 8
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 5 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
30.7%
Consistency (pass^4) · rank 8 of 8
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 26 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
89.7%
Task success (pass^1) · rank 5 of 8
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 5 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
57.2%
Task success (pass^1) · rank 8 of 8
- Unit
- % of tasks, higher is better
- Configuration
- gpt-5.2
- Measured
- 26 May 2026
- Not shown
- Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
76.0%
Requested-length error · rank 334 of 370
- Unit
- % off the requested word count, lower is better
- Configuration
- GPT-5.2 (high)
- Measured
- 4 Mar 2026
- Not shown
- Not other format limits such as character counts or bullet counts.
103.0%
Requested-length error · rank 350 of 370
- Unit
- % off the requested word count, lower is better
- Configuration
- gpt-5.2
- Measured
- 4 Mar 2026
- Not shown
- Not other format limits such as character counts or bullet counts.
93.0%
Requested-length error · rank 345 of 370
- Unit
- % off the requested word count, lower is better
- Configuration
- gpt-5.2
- Measured
- 4 Mar 2026
- Not shown
- Not other format limits such as character counts or bullet counts.
112.0%
Requested-length error · rank 355 of 370
- Unit
- % off the requested word count, lower is better
- Configuration
- gpt-5.2
- Measured
- 4 Mar 2026
- Not shown
- Not other format limits such as character counts or bullet counts.
0.38
Style adherence · rank 62 of 370
- Unit
- score from 0 to 1, higher is better
- Configuration
- GPT-5.2 (high)
- Measured
- 4 Mar 2026
- Not shown
- Not brand voice on your own examples: UGI's prompts are private and lean towards creative writing.
0.40
Style adherence · rank 34 of 370
- Unit
- score from 0 to 1, higher is better
- Configuration
- gpt-5.2
- Measured
- 4 Mar 2026
- Not shown
- Not brand voice on your own examples: UGI's prompts are private and lean towards creative writing.
0.37
Style adherence · rank 90 of 370
- Unit
- score from 0 to 1, higher is better
- Configuration
- gpt-5.2
- Measured
- 4 Mar 2026
- Not shown
- Not brand voice on your own examples: UGI's prompts are private and lean towards creative writing.
0.38
Style adherence · rank 65 of 370
- Unit
- score from 0 to 1, higher is better
- Configuration
- gpt-5.2
- Measured
- 4 Mar 2026
- Not shown
- Not brand voice on your own examples: UGI's prompts are private and lean towards creative writing.
49.4
Writing score · rank 159 of 370
- Unit
- score out of 100, higher is better
- Configuration
- GPT-5.2 (high)
- Measured
- 4 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.
43.9
Writing score · rank 185 of 370
- Unit
- score out of 100, higher is better
- Configuration
- gpt-5.2
- Measured
- 4 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.
44.1
Writing score · rank 182 of 370
- Unit
- score out of 100, higher is better
- Configuration
- gpt-5.2
- Measured
- 4 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.
39.2
Writing score · rank 220 of 370
- Unit
- score out of 100, higher is better
- Configuration
- gpt-5.2
- Measured
- 4 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.
100.0%
Answer rate · rank 1 of 108
- Unit
- % of documents, higher is better
- Configuration
- GPT-5.2 (high)
- 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.
100.0%
Answer rate · rank 1 of 108
- Unit
- % of documents, higher is better
- Configuration
- gpt-5.2
- 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.
10.8%
Hallucination rate · rank 68 of 108
- Unit
- % of summaries, lower is better
- Configuration
- GPT-5.2 (high)
- 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.
8.4%
Hallucination rate · rank 43 of 108
- Unit
- % of summaries, lower is better
- Configuration
- gpt-5.2
- 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.
