Qwen3-Next-80B-A3B-Thinking

MERA Created at 05.03.2026 00:40
0.71
The overall result
24
Place in the rating
In the top by tasks:
6
ruMultiAr
The result on the task is higher than human
The task is one of the main ones
Weak tasks:
20
RWSD
360
PARus
105
RCB
33
ruEthics
203
MultiQ
39
ruWorldTree
34
ruOpenBookQA
124
CheGeKa
25
ruMMLU
128
ruHateSpeech
106
ruDetox
86
ruHHH
20
ruTiE
23
ruHumanEval
38
USE
21
MathLogicQA
40
SimpleAr
23
LCS
70
BPS
22
MaMuRAMu
31
ruCodeEval
+17
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.442 Accuracy
RCB 0.566 / 0.438 Accuracy F1 macro
USE 0.414 Grade norm
RWSD 0.765 Accuracy
PARus 0.682 Accuracy
ruTiE 0.896 Accuracy
MultiQ 0.415 / 0.312 F1 Exact match
CheGeKa 0.306 / 0.257 F1 Exact match
ruModAr 0.998 Exact match
MaMuRAMu 0.874 Accuracy
ruMultiAr 1.0 Exact match
ruCodeEval 0.511 / 0.772 / 0.829 Pass@k
MathLogicQA 0.99 Accuracy
ruWorldTree 0.987 / 0.794 Accuracy F1 macro
ruOpenBookQA 0.94 / 0.761 Accuracy F1 macro

Evaluation on open tasks:

Go to the ratings by subcategory

The table will scroll to the left

Task name Result Metric
BPS 0.986 Accuracy
ruMMLU 0.82 Accuracy
SimpleAr 0.997 Exact match
ruHumanEval 0.617 / 0.838 / 0.866 Pass@k
ruHHH 0.843
ruHateSpeech 0.792
ruDetox 0.287
ruEthics
Correct God Ethical
Virtue 0.433 0.473 0.552
Law 0.464 0.488 0.553
Moral 0.489 0.517 0.589
Justice 0.416 0.445 0.494
Utilitarianism 0.374 0.416 0.471

Information about the submission

Mera version
v1.2.0
Torch Version
2.9.1
The version of the codebase
7030e16
CUDA version
12.8
Precision of the model weights
bfloat16
Seed
1234
Batch
1
Transformers version
4.56.1
The number of GPUs and their type
8 x NVIDIA A100-SXM4-80GB
Architecture
local-chat-completions

Team:

MERA

Name of the ML model:

Qwen3-Next-80B-A3B-Thinking

Model size

80.0B

Model type:

Opened

SFT

MoE

Additional links:

https://arxiv.org/abs/2505.09388

Architecture description:

Qwen3-Next-80B-A3B-Thinking is a large decoder-only transformer language model from the Qwen3 family with approximately 80 billion parameters. The model is designed for advanced reasoning tasks and generates intermediate reasoning steps before producing the final answer. It supports long-context inputs and improved reasoning capabilities compared to earlier Qwen models.

Description of the training:

The model follows a multi-stage training pipeline including large-scale pretraining followed by post-training stages. Post-training includes instruction tuning and reinforcement-learning-based alignment to improve reasoning, instruction following, and response quality.

Pretrain data:

The model was pretrained on a large multilingual corpus of approximately 36 trillion tokens covering 119 languages. It was further post-trained on instruction-following and reasoning-oriented datasets to improve instruction following and reasoning performance.

License:

Apache License 2.0

Inference parameters

System prompt:
Реши задачу по инструкции ниже. Не давай никаких объяснений и пояснений к своему ответу. Не пиши ничего лишнего. Пиши только то, что указано в инструкции. Если по инструкции нужно решить пример, то напиши только числовой ответ без хода решения и пояснений. Если по инструкции нужно вывести букву, цифру или слово, выведи только его. Если по инструкции нужно выбрать один из вариантов ответа и вывести букву или цифру, которая ему соответствует, то выведи только эту букву или цифру, не давай никаких пояснений, не добавляй знаки препинания, только 1 символ в ответе. Если по инструкции нужно дописать код функции на языке Python, пиши сразу код, соблюдая отступы так, будто ты продолжаешь функцию из инструкции, не давай пояснений, не пиши комментарии, используй только аргументы из сигнатуры функции в инструкции, не пробуй считывать данные через функцию input. Не извиняйся, не строй диалог. Выдавай только ответ и ничего больше.

Description of the template:
{%- if tools %} {{- '<|im_start|>system \n' }} {%- if messages[0].role == 'system' %} {{- messages[0].content + ' \n \n' }} {%- endif %} {{- "# Tools \n \nYou may call one or more functions to assist with the user query. \n \nYou are provided with function signatures within <tools></tools> XML tags: \n<tools>" }} {%- for tool in tools %} {{- " \n" }} {{- tool | tojson }} {%- endfor %} {{- " \n</tools> \n \nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags: \n<tool_call> \n{"name": <function-name>, "arguments": <args-json-object>} \n</tool_call><|im_end|> \n" }} {%- else %} {%- if messages[0].role == 'system' %} {{- '<|im_start|>system \n' + messages[0].content + '<|im_end|> \n' }} {%- endif %} {%- endif %} {%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %} {%- for message in messages[::-1] %} {%- set index = (messages|length - 1) - loop.index0 %} {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %} {%- set ns.multi_step_tool = false %} {%- set ns.last_query_index = index %} {%- endif %} {%- endfor %} {%- for message in messages %} {%- if message.content is string %} {%- set content = message.content %} {%- else %} {%- set content = '' %} {%- endif %} {%- if (message.role == "user") or (message.role == "system" and not loop.first) %} {{- '<|im_start|>' + message.role + ' \n' + content + '<|im_end|>' + ' \n' }} {%- elif message.role == "assistant" %} {%- set reasoning_content = '' %} {%- if message.reasoning_content is string %} {%- set reasoning_content = message.reasoning_content %} {%- else %} {%- if '</think>' in content %} {%- set reasoning_content = content.split('</think>')[0].rstrip(' \n').split('<think>')[-1].lstrip(' \n') %} {%- set content = content.split('</think>')[-1].lstrip(' \n') %} {%- endif %} {%- endif %} {%- if loop.index0 > ns.last_query_index %} {%- if loop.last or (not loop.last and reasoning_content) %} {{- '<|im_start|>' + message.role + ' \n<think> \n' + reasoning_content.strip(' \n') + ' \n</think> \n \n' + content.lstrip(' \n') }} {%- else %} {{- '<|im_start|>' + message.role + ' \n' + content }} {%- endif %} {%- else %} {{- '<|im_start|>' + message.role + ' \n' + content }} {%- endif %} {%- if message.tool_calls %} {%- for tool_call in message.tool_calls %} {%- if (loop.first and content) or (not loop.first) %} {{- ' \n' }} {%- endif %} {%- if tool_call.function %} {%- set tool_call = tool_call.function %} {%- endif %} {{- '<tool_call> \n{"name": "' }} {{- tool_call.name }} {{- '", "arguments": ' }} {%- if tool_call.arguments is string %} {{- tool_call.arguments }} {%- else %} {{- tool_call.arguments | tojson }} {%- endif %} {{- '} \n</tool_call>' }} {%- endfor %} {%- endif %} {{- '<|im_end|> \n' }} {%- elif message.role == "tool" %} {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %} {{- '<|im_start|>user' }} {%- endif %} {{- ' \n<tool_response> \n' }} {{- content }} {{- ' \n</tool_response>' }} {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %} {{- '<|im_end|> \n' }} {%- endif %} {%- endif %} {%- endfor %} {%- if add_generation_prompt %} {{- '<|im_start|>assistant \n' }} {%- if enable_thinking is defined and enable_thinking is false %} {{- '<think> \n \n</think> \n \n' }} {%- endif %} {%- endif %}

Ratings by subcategory

Metric: Grade Norm
Model, team 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 8_0 8_1 8_2 8_3 8_4
Qwen3-Next-80B-A3B-Thinking
MERA
0.433 0.367 0.8 0.133 0.467 0.8 0.567 - 0.067 0 0.333 0.333 0.5 0.267 0.4 0.333 0.4 0.5 0.5 0.333 0.267 0.367 0.2 0.433 0.467 0.392 0.333 0.5 0.6 0.667 0.8
Model, team Honest Helpful Harmless
Qwen3-Next-80B-A3B-Thinking
MERA
0.836 0.746 0.948
Model, team Anatomy Virology Astronomy Marketing Nutrition Sociology Management Philosophy Prehistory Human aging Econometrics Formal logic Global facts Jurisprudence Miscellaneous Moral disputes Business ethics Biology (college) Physics (college) Human Sexuality Moral scenarios World religions Abstract algebra Medicine (college) Machine learning Medical genetics Professional law PR Security studies Chemistry (школьная) Computer security International law Logical fallacies Politics Clinical knowledge Conceptual_physics Math (college) Biology (high school) Physics (high school) Chemistry (high school) Geography (high school) Professional medicine Electrical engineering Elementary mathematics Psychology (high school) Statistics (high school) History (high school) Math (high school) Professional accounting Professional psychology Computer science (college) World history (high school) Macroeconomics Microeconomics Computer science (high school) European history Government and politics
Qwen3-Next-80B-A3B-Thinking
MERA
0.756 0.548 0.967 0.872 0.83 0.866 0.835 0.814 0.864 0.749 0.746 0.881 0.6 0.833 0.891 0.737 0.8 0.951 0.956 0.847 0.632 0.842 0.93 0.792 0.795 0.93 0.613 0.694 0.771 0.73 0.85 0.934 0.791 0.919 0.815 0.91 0.93 0.939 0.907 0.921 0.859 0.941 0.841 0.96 0.94 0.866 0.902 0.97 0.791 0.808 0.94 0.886 0.905 0.966 0.93 0.867 0.938
Model, team SIM FL STA
Qwen3-Next-80B-A3B-Thinking
MERA
0.581 0.631 0.813
Model, team Anatomy Virology Astronomy Marketing Nutrition Sociology Managment Philosophy Pre-History Gerontology Econometrics Formal logic Global facts Jurisprudence Miscellaneous Moral disputes Business ethics Bilology (college) Physics (college) Human sexuality Moral scenarios World religions Abstract algebra Medicine (college) Machine Learning Genetics Professional law PR Security Chemistry (college) Computer security International law Logical fallacies Politics Clinical knowledge Conceptual physics Math (college) Biology (high school) Physics (high school) Chemistry (high school) Geography (high school) Professional medicine Electrical Engineering Elementary mathematics Psychology (high school) Statistics (high school) History (high school) Math (high school) Professional Accounting Professional psychology Computer science (college) World history (high school) Macroeconomics Microeconomics Computer science (high school) Europe History Government and politics
Qwen3-Next-80B-A3B-Thinking
MERA
0.778 0.921 0.833 0.759 0.921 0.81 0.759 0.702 0.904 0.8 0.808 0.85 0.642 0.922 0.86 0.827 0.766 0.844 0.912 0.842 0.842 0.949 0.956 0.941 0.956 0.909 0.897 0.737 0.93 0.889 0.889 0.923 0.875 0.965 0.864 0.911 0.933 0.911 0.912 0.923 0.943 0.952 0.867 0.978 0.931 0.911 0.948 0.909 0.892 1 0.889 0.957 0.848 0.766 0.837 0.836 0.944
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Qwen3-Next-80B-A3B-Thinking
MERA
0.433 0.464 0.489 0.416 0.374
Model, team Virtue Law Moral Justice Utilitarianism
Qwen3-Next-80B-A3B-Thinking
MERA
0.473 0.488 0.517 0.445 0.416
Model, team Virtue Law Moral Justice Utilitarianism
Qwen3-Next-80B-A3B-Thinking
MERA
0.552 0.553 0.589 0.494 0.471
Model, team Women Men LGBT Nationalities Migrants Other
Qwen3-Next-80B-A3B-Thinking
MERA
0.824 0.6 0.824 0.865 1 0.77