[whatmodel.sh](../../README.md) / MiniMax-M1 (80k) # MiniMax-M1 (80k) MiniMax · released 2025-06 · 456B params (45.9B active) · apache-2.0 **score 62.0** ±1.2 · #182 of 732 · from 13 benchmark results across 2 sources Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only. ## What it's good at ● marks MiniMax-M1 (80k) among every model on that task or test, worst to best. 
task            vs the others          rank  counts  tests
agent tasks     ······●·····       #1 of 1*    8.9%      3
coding          ······●·····       #1 of 1*     13%      2
chat & writing  ▁▁▁▃▆▄▆█●▇▇▃   #151 of 402*     16%      1
math            ······●·····        #1 of 1     25%      3
agent tasks · #1 of 1 · 8.9% of its score 
test                                result  vs the others in its run   difficulty      counts  why
SWE-bench Verified (maker)           56.0%  ▁▁▁▁▁▁▂●█▆▁▁ beats 47/138  hard · avg 23%    8.9%  sharp · at its level · small run
τ-bench Airline (GPT-4.1 user) (m…   62.0%  █··········● beats 1/1                          —  not counted yet: no other model in the …
τ-bench Retail (GPT-4.1 user) (ma…   63.5%  ●··········█ beats 0/1                          —  not counted yet: no other model in the …
coding · #1 of 1 · 13% of its score 
test                                result  vs the others in its run   difficulty        counts  why
SWE-bench Verified (maker)           56.0%  ▁▁▁▁▁▁▂●█▆▁▁ beats 47/138  hard · avg 23%      8.9%  sharp · at its level · small run
LiveCodeBench 2024-08–2025-05 (ve…   65.0%  ▃···●··▅··▃▃ beats 3/7     medium · avg 43%    3.8%  at its level · small run
chat & writing · #151 of 402 · 16% of its score Beats [INTELLECT-3](intellect-3.md), [MiniMax-M2](minimax-m2.md), [Nemotron 3 Super 120B-A12B](nemotron-3-super-120b-a12b.md) here, though they rank higher overall. Trails [LFM2.5 1.2B Thinking](lfm2-5-1-2b-thinking.md), [LFM2.5 8B-A1B](lfm2-5-8b-a1b.md), [LFM2.5 2.6B](lfm2-5-2-6b.md) here, though they rank lower overall. 
test            result  vs the others in its run    difficulty  counts  why
LMArena (text)  1363.8  ▁▁▁▂▃▃▃▆●▇█▃ beats 208/385  avg 1260       16%  blunt
math · #1 of 1 · 25% of its score 
test               result  vs the others in its run   difficulty           counts  why
AIME 2025 (maker)   76.9%  ▁·▁▁▁▂▅▆●▇▅▁ beats 77/133  medium · avg 41%        10%  sharp · small run
AIME 2024 (maker)   86.0%  ▁▁▁▂▁▃▁▅█●█▃ beats 61/85   medium · avg 56%       9.2%  sharp · small run
MATH-500 (maker)    96.8%  ▁▃▂▃▄▇▅▇●▄▂▁ beats 65/85   very easy · avg 89%    5.2%  maxed out · small run
other tests (7) 
test                          result  vs the others in its run    difficulty            counts  why
GPQA Diamond (maker)           70.0%  ▁▁▁▁▂▄▅▅●██▄ beats 156/322  hard · avg 41%           13%  sharp · at its level · small run
LongBench v2 (maker)           61.5%  ▂·▁·▂▃▁▄▄▃●▄ beats 28/33    hard · avg 39%          6.5%  at its level · small run
Humanity's Last Exam (maker)    8.4%  ▁▁▁●▆▆▇██▆▁▃ beats 13/111   very hard · avg 3.1%    6.4%  small run
ZebraLogic (maker)             86.8%  █·▁▁▁▇·●▄▁▄▁ beats 14/24    medium · avg 59%        5.8%  small run
MMLU-Pro (maker)               81.1%  ▁▁·▁▁▁▃▂▄▅●▁ beats 131/188  medium · avg 59%        5.7%  small run
SimpleQA (maker)               18.5%  ▄▄▄▆█▆●██·▂█ beats 19/32    very hard · avg 7.5%    5.6%  small run
MultiChallenge (maker)         44.7%  ▁·▁·▁▃●▅▃▂█▅ beats 9/30     hard · avg 27%          3.6%  at its level · small run
## How the score was worked out Each result implies a score on its own, after adjusting for how hard that test is and who ran it. The model's score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its *counts* column above; here it is added up by who ran the tests. [Method](../method.md) 
source                                    results  share of the score
MiniMaxAI/MiniMax-M1-80k model card (su…       12   83% ▇▇▇▇▇▇▇▇▇▇
LMArena leaderboard dataset                     1   16% ▇▇
prior: pull toward the middle                        1% ▇
weighted average                                   = 62.0
Licences: MiniMaxAI/MiniMax-M1-80k model card ([apache-2.0](../../licenses/MiniMaxAI--MiniMax-M1-80k.txt)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/))
Not counted (2) 
source                  test                            score  why
MiniMaxAI/MiniMax-M1-…  τ-bench Airline (GPT-4.1 user)  62.0%  too few models have it yet
MiniMaxAI/MiniMax-M1-…  τ-bench Retail (GPT-4.1 user)   63.5%  too few models have it yet
## How to run it ### Through an API api (Novita AI) · $0.96/M blended · $0.55 in / $2.2 out per 1M tokens · [novita.ai](https://novita.ai/models) [price source](https://github.com/BerriAI/litellm/blob/c19ce71bcdad234ca87a7af10cee45375ce811a8/model_prices_and_context_window.json): USD per 1M text tokens, standard uncached input and output. Model endpoint minimaxai/minimax-m1-80k. Product: Serverless chat completions. API model: minimaxai/minimax-m1-80k. Settings: service tier=standard, batch=no, cache=uncached input. Additional/excluded charges: Tools, cache storage, priority service and other modalities excluded.. Currency: USD. · [context source](https://github.com/BerriAI/litellm/blob/c19ce71bcdad234ca87a7af10cee45375ce811a8/model_prices_and_context_window.json): provider maximum input tokens; not an intrinsic model maximum or legacy max\_tokens ### Other ways to run it 1. Run it as published (BF16, needs ~912.21 GB, from [huggingface.co/MiniMaxAI](https://huggingface.co/MiniMaxAI/MiniMax-M1-80k/blob/8d1494b1a260e22040d5b9b2eb332eb44500b34d/docs/transformers_deployment_guide.md)) ```sh pip install transformers torch accelerate && python -c "from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig; p='MiniMaxAI/MiniMax-M1-80k'; m=AutoModelForCausalLM.from_pretrained(p, device_map='auto', trust_remote_code=True); t=AutoTokenizer.from_pretrained(p, trust_remote_code=True); x=t(t.apply_chat_template([{'role': 'user', 'content': [{'type': 'text', 'text': 'Do you have mayonnaise recipes?'}]}], tokenize=False, add_generation_prompt=True), return_tensors='pt').to(m.device); o=m.generate(**x, generation_config=GenerationConfig(max_new_tokens=20, eos_token_id=t.eos_token_id, use_cache=True)); print(t.batch_decode([o[0][x.input_ids.shape[1]:]], skip_special_tokens=True)[0])" ``` Model repository and documentation: [huggingface.co/MiniMaxAI](https://huggingface.co/MiniMaxAI/MiniMax-M1-80k) [interactive version](https://whatmodel.sh/model/minimax-m1) · [JSON](https://whatmodel.sh/api?format=json&all=1)