FlagEval Debate
Models debate a topic in Chinese. The leaderboard estimates their ratings from comparisons and reports uncertainty.
The official ratings feed lists 21 models.
- Released
- September 2024
- Kind
- Overall chat quality
- Level
- Everyday
- Scored
- Ranked by votes
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Preference
- Language
- Chinese
- Interaction
- Multi-agent
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Agents
- Strongest model tested
- o1-mini #142 on the capability index
- Last model added
- December 2024
- Scores refreshed
- November 21st, 2026
o1-mini leads with 1,149 among models released by December 2024 o1-mini leads with 1,149 among models released by December 2024
Best score per entry from the benchmark's own leaderboard, read on November 21st, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in December 2024, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1,149 o1-mini
- 1,135 Doubao-pro-32k-240828
- 1,132 Nanbeige2-Turbo-0611
- 1,125 GLM-4-Plus
- 1,113
Yi-Lightning
- 1,095
DeepSeek-V2.5
- 1,091 Hunyuan-Turbo
- 1,075 o1-preview
- 1,070 GPT-4o
- 1,045
Gemini 1.5 Pro
- 1,006 SenseChat-5
- 1,005 GPT-4o
- 997 Qwen-Max-0919
- 993 GPT-4o mini
- 992 GLM-4
- 977 Moonshot-v1-8k
- 945 Yi-34B-Chat
- 936
Yi-Large
- 930 Baichuan2-53B
- 895 Baichuan 4
- 864 Doubao-pro-4k
Bars start at 800.
- 1 o1-mini 1,149
- 2 Doubao-pro-32k-240828 1,135
- 3 Nanbeige2-Turbo-0611 1,132
- 4 GLM-4-Plus 1,125
- 5
Yi-Lightning 1,113
- 6
DeepSeek-V2.5 Sep 2024 1,095
- 7 Hunyuan-Turbo 1,091
- 8 o1-preview 1,075
- 9 GPT-4o Aug 2024 1,070
- 10
Gemini 1.5 Pro Sep 2024 1,045
- 11 SenseChat-5 1,006
- 12 GPT-4o Nov 2024 1,005
- 13 Qwen-Max-0919 997
- 14 GPT-4o mini 993
- 15 GLM-4 992
- 16 Moonshot-v1-8k 977
- 17 Yi-34B-Chat 945
- 18
Yi-Large 936
- 19 Baichuan2-53B 930
- 20 Baichuan 4 895
- 21 Doubao-pro-4k 864
Source: Official FlagEval Debate leaderboard
Bars start at 900.
- 1 o1-mini 1,149
- 2
Yi-Lightning 1,113
- 3
DeepSeek-V2.5 Sep 2024 1,095
- 4 o1-preview 1,075
- 5 GPT-4o Aug 2024 1,070
- 6
Gemini 1.5 Pro Sep 2024 1,045
- 7 GPT-4o Nov 2024 1,005
- 8 GPT-4o mini 993
- 9 GLM-4 992
- 10
Yi-Large 936
Source: Official FlagEval Debate leaderboard
| # | Entry | Score ± | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | o1-mini | 1,149 1136.1 to 1160.8 | Sep 2024 | Official leaderboard | |
| 2 | Doubao-pro-32k-240828 | 1,135 1124.6 to 1145.4 | Official leaderboard | ||
| 3 | Nanbeige2-Turbo-0611 | 1,132 ±11.4 | Official leaderboard | ||
| 4 | GLM-4-Plus | 1,125 1113.5 to 1139.1 | Official leaderboard | ||
| 5 | | 1,113 1099.4 to 1123.4 | Dec 2024 | Official leaderboard | |
| 6 | | 1,095 ±11.6 | Sep 2024 | Official leaderboard | |
| 7 | Hunyuan-Turbo | 1,091 ±12.6 | Official leaderboard | ||
| 8 | o1-preview | 1,075 ±9.8 | Sep 2024 | Official leaderboard | |
| 9 | GPT-4o Aug 2024 | 1,070 ±12.1 | Aug 2024 | Official leaderboard | |
| 10 | | 1,045 1037.2 to 1057.4 | Sep 2024 | Official leaderboard | |
| 11 | SenseChat-5 | 1,006 ±11.3 | Official leaderboard | ||
| 12 | GPT-4o Nov 2024 | 1,005 998.1 to 1015.7 | Nov 2024 | Official leaderboard | |
| 13 | Qwen-Max-0919 | 997 988 to 1009.2 | Official leaderboard | ||
| 14 | GPT-4o mini | 993 984.2 to 1004 | Jul 2024 | Official leaderboard | |
| 15 | GLM-4 | 992 981.3 to 1007 | Jan 2024 | Official leaderboard | |
| 16 | Moonshot-v1-8k | 977 ±10.3 | Official leaderboard | ||
| 17 | Yi-34B-Chat | 945 ±12.1 | Official leaderboard | ||
| 18 | | 936 924.9 to 948.3 | May 2024 | Official leaderboard | |
| 19 | Baichuan2-53B | 930 916.9 to 944.3 | Official leaderboard | ||
| 20 | Baichuan 4 | 895 883.7 to 905.1 | Official leaderboard | ||
| 21 | Doubao-pro-4k | 864 851.5 to 875.4 | Official leaderboard |
| # | Model | Score ± | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | o1-mini | 1,149 1136.1 to 1160.8 | Sep 2024 | Official leaderboard | |
| 2 | | 1,113 1099.4 to 1123.4 | Dec 2024 | Official leaderboard | |
| 3 | | 1,095 ±11.6 | Sep 2024 | Official leaderboard | |
| 4 | o1-preview | 1,075 ±9.8 | Sep 2024 | Official leaderboard | |
| 5 | GPT-4o Aug 2024 | 1,070 ±12.1 | Aug 2024 | Official leaderboard | |
| 6 | | 1,045 1037.2 to 1057.4 | Sep 2024 | Official leaderboard | |
| 7 | GPT-4o Nov 2024 | 1,005 998.1 to 1015.7 | Nov 2024 | Official leaderboard | |
| 8 | GPT-4o mini | 993 984.2 to 1004 | Jul 2024 | Official leaderboard | |
| 9 | GLM-4 | 992 981.3 to 1007 | Jan 2024 | Official leaderboard | |
| 10 | | 936 924.9 to 948.3 | May 2024 | Official leaderboard |
3 models have held number one on the official leaderboard since its release in September 2024
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official FlagEval Debate leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official FlagEval Debate leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Instructions and chat
- IFEval In use Models answer about 500 prompts containing verifiable instructions such as word limits or required keywords. Code checks each instruction for strict and loose accuracy.
Gemma 4 31B IT 98.9%
- IFBench In use 58 new out-of-domain verifiable output constraints. Scored by programmatic constraint checks (strict or loose). Flint 89.2%
- MultiChallenge In use Multi-turn conversations testing instruction retention, inference memory, versioned editing and self-coherence. Graded by LLM judge with rubrics. GPT-6.1 Sol 82%
- Multi-IF In use Models follow verifiable instructions over three-turn conversations in eight languages, extending IFEval. Scored by instruction accuracy at each turn. LFM3-1.7B 89.1%
- AdvancedIF In use Models respond to over 1,600 prompts with complex, multi-turn, and system-level instructions. A judge model checks each response against expert-written rubrics.
MAI-Thinking-1 85%
- MM-IFEval In use Multimodal models follow image-grounded instructions with exact output constraints. Checks mix rule-based verification and a judge model. GPT-4o 64.6%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 23 months ago, in December 2024, for Yi-Lightning. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (21 entries).
- Pass Models scored 10 models we track, plus 11 other entries on its leaderboard.
- Warning Newest score The newest came 23 months ago, in December 2024, for Yi-Lightning.
- Note Frontier models None of the top 10 has a score. The strongest tested is o1-mini (#142).
- Note Headroom Scored as a rating, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on flageval.baai.ac.cn: 10 rows for models we track and 11 other entries at the last read.
- Pass Last read November 21st, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.