Qwen3-30B-A3B-Thinking-2507

MERA Created at 04.03.2026 14:50
0.747
The overall result
17
Place in the rating
In the top by tasks:
7
ruEthics
10
ruHateSpeech
7
MathLogicQA
The result on the task is higher than human
The task is one of the main ones
4
ruMultiAr
The result on the task is higher than human
The task is one of the main ones
10
SimpleAr
+1
Hide
Weak tasks:
28
RWSD
194
PARus
138
MultiQ
44
ruWorldTree
33
ruOpenBookQA
134
CheGeKa
34
ruMMLU
26
ruDetox
21
ruHHH
22
ruHumanEval
29
USE
185
BPS
38
MaMuRAMu
33
ruCodeEval
+10
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.556 Accuracy
RCB 0.6 / 0.58 Accuracy F1 macro
USE 0.504 Grade norm
RWSD 0.735 Accuracy
PARus 0.852 Accuracy
ruTiE 0.909 Accuracy
MultiQ 0.469 / 0.316 F1 Exact match
CheGeKa 0.259 / 0.195 F1 Exact match
ruModAr 0.998 Exact match
MaMuRAMu 0.85 Accuracy
ruMultiAr 1.0 Exact match
ruCodeEval 0.499 / 0.735 / 0.787 Pass@k
MathLogicQA 0.997 Accuracy
ruWorldTree 0.985 / 0.985 Accuracy F1 macro
ruOpenBookQA 0.94 / 0.94 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 0.958 Accuracy
ruMMLU 0.804 Accuracy
SimpleAr 1.0 Exact match
ruHumanEval 0.623 / 0.776 / 0.811 Pass@k
ruHHH 0.888
ruHateSpeech 0.921
ruDetox 0.352
ruEthics
Correct God Ethical
Virtue 0.585 0.517 0.657
Law 0.634 0.53 0.639
Moral 0.642 0.553 0.688
Justice 0.553 0.485 0.613
Utilitarianism 0.522 0.445 0.542

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
4 x NVIDIA A100-SXM4-80GB
Architecture
local-chat-completions

Team:

MERA

Name of the ML model:

Qwen3-30B-A3B-Thinking-2507

Model size

30.0B

Model type:

Opened

SFT

MoE

Additional links:

https://arxiv.org/abs/2505.09388

Architecture description:

Qwen3-30B-A3B-Thinking-2507 is a decoder-only transformer language model from the Qwen3 family with approximately 30 billion parameters. The model is optimized for complex 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 the standard Qwen3 multi-stage training pipeline consisting of large-scale pretraining followed by post-training. Post-training includes instruction tuning and reinforcement-learning-based alignment to improve reasoning ability, instruction following, and response quality.

Pretrain data:

The model was pretrained on a large multilingual corpus of approximately 36 trillion tokens covering 119 languages, including web documents, books, code repositories, and technical materials. Additional post-training datasets are used to enhance reasoning and instruction-following capabilities.

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
Qwen3-30B-A3B-Thinking-2507
MERA
0.467 0.567 0.867 0.167 0.467 0.633 0.467 - 0.1 0.133 0.1 0.433 0.367 0.467 0.2 0.633 0.433 0.333 0.4 0.4 0.5 0.733 0.467 0.567 0.6 0.692 0.4 0.667 0.8 0.633 0.733
Model, team Honest Helpful Harmless
Qwen3-30B-A3B-Thinking-2507
MERA
0.852 0.864 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
Qwen3-30B-A3B-Thinking-2507
MERA
0.748 0.536 0.921 0.91 0.85 0.836 0.845 0.762 0.84 0.717 0.754 0.897 0.56 0.806 0.871 0.725 0.77 0.944 0.967 0.855 0.622 0.86 0.94 0.792 0.848 0.97 0.553 0.639 0.788 0.72 0.82 0.835 0.761 0.818 0.845 0.927 0.95 0.939 0.907 0.921 0.879 0.923 0.834 0.942 0.934 0.898 0.892 0.985 0.748 0.754 0.91 0.873 0.918 0.966 0.94 0.848 0.902
Model, team SIM FL STA
Qwen3-30B-A3B-Thinking-2507
MERA
0.729 0.654 0.761
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
Qwen3-30B-A3B-Thinking-2507
MERA
0.733 0.871 0.783 0.75 0.934 0.793 0.724 0.719 0.808 0.8 0.769 0.85 0.65 0.884 0.813 0.765 0.785 0.844 0.895 0.86 0.789 0.898 0.956 0.947 0.956 0.909 0.833 0.667 0.947 0.933 0.844 0.923 0.848 0.947 0.712 0.875 0.933 0.889 0.912 0.954 0.91 0.921 0.8 1 0.914 0.911 0.931 0.932 0.892 0.982 0.867 0.855 0.861 0.805 0.814 0.737 0.911
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Qwen3-30B-A3B-Thinking-2507
MERA
0.585 0.634 0.642 0.553 0.522
Model, team Virtue Law Moral Justice Utilitarianism
Qwen3-30B-A3B-Thinking-2507
MERA
0.517 0.53 0.553 0.485 0.445
Model, team Virtue Law Moral Justice Utilitarianism
Qwen3-30B-A3B-Thinking-2507
MERA
0.657 0.639 0.688 0.613 0.542
Model, team Women Men LGBT Nationalities Migrants Other
Qwen3-30B-A3B-Thinking-2507
MERA
0.944 0.8 0.941 0.946 0.857 0.934