Skip to content
Weights Not Walls
Menu

MiniMaxAI

MiniMax-M2.5

Reviewed
Size
229B parameterscounted from the safetensors header on the Hub
Licence
custom, read it, MiniMax Model LicenseOur reading, check the licence: A custom licence we have not read for this model: other makers use the same name for different texts. Read the model's own licence file before you use it.Licence text
On the Hub since
12 Feb 2026the date the repo was created
Context length
192k tokensfrom the model's config
Kind
Base model (no base model declared on the Hub)MiniMaxM2ForCausalLM
On Hugging Face
MiniMaxAI/MiniMax-M2.5967k downloads in the last 30 days, 1.5k likes

Scores

7 independent results on 6 eval sets
ARC-AGIReported

Epoch's name: MiniMax-M2.5. Best closed model on this set: claude-fable-5_max, 98.5% (Anthropic). All 318 results

ARC-AGI-2Reported

Epoch's name: MiniMax-M2.5. Best closed model on this set: gpt-6-astra_max, 95.0% (OpenAI). All 299 results

CL-benchReported

Epoch's name: MiniMax-M2.5. Best closed model on this set: gpt-5.4-2026-03-05_xhigh, 27.9% (OpenAI). All 23 results

Epoch's name: MiniMax-M2.5. Best closed model on this set: gpt-5.5_high, 22.2% (OpenAI). All 17 results

ProofBenchReported

Epoch's name: MiniMax-M2.5. Best closed model on this set: claude-fable-5-1_max, 100.0% (Anthropic). All 85 results

  • Agent: cchuter42.7%±2.812 Feb 2026, tbench.ai
  • Agent: Terminus 242.2%±2.612 Feb 2026, tbench.ai

Epoch's name: MiniMax-M2.5. Best closed model on this set: gpt-5.5_unknown, 84.7% (OpenAI). All 202 results

Claimed by the maker Claimed

The model card has no claimed scores in a machine-readable form (the card's model-index), so none are shown. The maker may give scores in the card text: read the card.

Other views, which we link to and do not copy: look this model up on Artificial Analysis and the Arena leaderboard.

Independent scores from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Each row links its own source. What the badges mean.

Run it yourself

file sizes read from the Hub 19 Nov 2026

Memory needed = weights + context cache + 0.50 GB for the runtime. Context cache at 8k tokens: 1.9 GB, worked out from the model's config (MiniMaxAI/MiniMax-M2.5), at 16-bit precision. Choose a machine above to see a verdict for each format. Method

GGUF for llama.cpp, Ollama, LM Studio and others

unsloth/MiniMax-M2.5-GGUF (converted by unsloth), 5.6k downloads. Sizes are the real files on the Hub, all parts added up.

FormatSizeMemory neededGet it
UD-IQ3_XXS86.9 GBfile size, 3 parts89.3 GB86.9 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:UD-IQ3_XXS
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "UD-IQ3_XXS/MiniMax-M2.5-UD-IQ3_XXS-*-of-00003.gguf" --local-dir MiniMax-M2.5-GGUF
Q3_K_S91.9 GBfile size, 3 parts94.4 GB91.9 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:Q3_K_S
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "Q3_K_S/MiniMax-M2.5-Q3_K_S-*-of-00003.gguf" --local-dir MiniMax-M2.5-GGUF
UD-Q3_K_XL94.3 GBfile size, 4 parts96.8 GB94.3 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:UD-Q3_K_XL
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "UD-Q3_K_XL/MiniMax-M2.5-UD-Q3_K_XL-*-of-00004.gguf" --local-dir MiniMax-M2.5-GGUF
Q3_K_M101.8 GBfile size, 4 parts104.2 GB101.8 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:Q3_K_M
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "Q3_K_M/MiniMax-M2.5-Q3_K_M-*-of-00004.gguf" --local-dir MiniMax-M2.5-GGUF
IQ4_XS113.5 GBfile size, 4 parts116.0 GB113.5 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:IQ4_XS
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "IQ4_XS/MiniMax-M2.5-IQ4_XS-*-of-00004.gguf" --local-dir MiniMax-M2.5-GGUF
MXFP4_MOE115.3 GBfile size, 4 parts117.7 GB115.3 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:MXFP4_MOE
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "MXFP4_MOE/MiniMax-M2.5-MXFP4_MOE-*-of-00004.gguf" --local-dir MiniMax-M2.5-GGUF
IQ4_NL120.1 GBfile size, 4 parts122.6 GB120.1 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:IQ4_NL
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "IQ4_NL/MiniMax-M2.5-IQ4_NL-*-of-00004.gguf" --local-dir MiniMax-M2.5-GGUF
Q4_0120.6 GBfile size, 4 parts123.0 GB120.6 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:Q4_0
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "Q4_0/MiniMax-M2.5-Q4_0-*-of-00004.gguf" --local-dir MiniMax-M2.5-GGUF
Q4_K_S121.1 GBfile size, 4 parts123.5 GB121.1 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:Q4_K_S
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "Q4_K_S/MiniMax-M2.5-Q4_K_S-*-of-00004.gguf" --local-dir MiniMax-M2.5-GGUF
UD-Q4_K_XL122.3 GBfile size, 4 parts124.7 GB122.3 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:UD-Q4_K_XL
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "UD-Q4_K_XL/MiniMax-M2.5-UD-Q4_K_XL-*-of-00004.gguf" --local-dir MiniMax-M2.5-GGUF
Q4_K_M128.8 GBfile size, 4 parts131.3 GB128.8 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:Q4_K_M
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "Q4_K_M/MiniMax-M2.5-Q4_K_M-*-of-00004.gguf" --local-dir MiniMax-M2.5-GGUF
Q4_1133.4 GBfile size, 4 parts135.8 GB133.4 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:Q4_1
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "Q4_1/MiniMax-M2.5-Q4_1-*-of-00004.gguf" --local-dir MiniMax-M2.5-GGUF
Q5_K_S146.7 GBfile size, 5 parts149.1 GB146.7 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:Q5_K_S
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "Q5_K_S/MiniMax-M2.5-Q5_K_S-*-of-00005.gguf" --local-dir MiniMax-M2.5-GGUF
UD-Q5_K_XL150.7 GBfile size, 5 parts153.1 GB150.7 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:UD-Q5_K_XL
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "UD-Q5_K_XL/MiniMax-M2.5-UD-Q5_K_XL-*-of-00005.gguf" --local-dir MiniMax-M2.5-GGUF
Q5_K_M151.2 GBfile size, 5 parts153.6 GB151.2 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:Q5_K_M
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "Q5_K_M/MiniMax-M2.5-Q5_K_M-*-of-00005.gguf" --local-dir MiniMax-M2.5-GGUF
Q6_K174.9 GBfile size, 5 parts177.3 GB174.9 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:Q6_K
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "Q6_K/MiniMax-M2.5-Q6_K-*-of-00005.gguf" --local-dir MiniMax-M2.5-GGUF
UD-Q6_K_XL180.9 GBfile size, 5 parts183.4 GB180.9 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:UD-Q6_K_XL
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "UD-Q6_K_XL/MiniMax-M2.5-UD-Q6_K_XL-*-of-00005.gguf" --local-dir MiniMax-M2.5-GGUF
Q8_0226.4 GBfile size, 6 parts228.9 GB226.4 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:Q8_0
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "Q8_0/MiniMax-M2.5-Q8_0-*-of-00006.gguf" --local-dir MiniMax-M2.5-GGUF
UD-Q8_K_XL243.4 GBfile size, 7 parts245.9 GB243.4 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:UD-Q8_K_XL
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "UD-Q8_K_XL/MiniMax-M2.5-UD-Q8_K_XL-*-of-00007.gguf" --local-dir MiniMax-M2.5-GGUF
BF16426.1 GBfile size, 10 parts428.5 GB426.1 + 1.9 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/MiniMax-M2.5-GGUF:BF16
Download
hf download unsloth/MiniMax-M2.5-GGUF --include "BF16/MiniMax-M2.5-BF16-*-of-00010.gguf" --local-dir MiniMax-M2.5-GGUF

MLX for Apple silicon

FormatSizeMemory neededRepo
MLX 6bitlmstudio-community/MiniMax-M2.5-MLX-6bit173.1 GBfile size175.5 GB173.1 + 1.9 + 0.50Download page
MLX 8bitlmstudio-community/MiniMax-M2.5-MLX-8bit226.3 GBfile size228.7 GB226.3 + 1.9 + 0.50Download page

Original weights size from the safetensors header, for vLLM, SGLang, Transformers

FormatSizeMemory neededRepo
F8_E4M3original weights214.3 GBfile size216.7 GB214.3 + 1.9 + 0.50
hf download MiniMaxAI/MiniMax-M2.5 --local-dir MiniMax-M2.5
Other conversions (19; sizes not read)

The Hub tags 64 quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.

Run it by API

Hugging Face Inference Providers, fetched 19 Nov 2026, 13:58 UTC

4 providers serve this model through the Hugging Face router. Prices are US dollars per million tokens. Precision is what the provider reports; a lower precision than the original weights can change answers. Live first, then cheapest output.

ProviderInOutContextPrecisionTokens/sFirst tokenUptime 24hToolsJSON
Venice$0.27$0.95193knot given580.9 s98.2%yesno
DigitalOcean$0.30$1.2064knot given710.6 s81.7%yesyes
novita$0.30$1.20200knot given460.8 s?yesno
featherless-ai???not given???not givennot given

Through the router, with a Hugging Face token (the router bills your Hugging Face account). Change the name after the colon to pick another provider.

curl https://router.huggingface.co/v1/chat/completions \ -H "Authorization: Bearer $HF_TOKEN" -H "Content-Type: application/json" \ -d '{"model": "MiniMaxAI/MiniMax-M2.5:venice", "messages": [{"role": "user", "content": "Hello"}]}'

Not built yet: one key for every open model, with BlockRun, and our own tested-quality score per provider. Until then, precision and measured speed are the best guide to a provider here.

Planned for this page Not built yet

Try it in private chat
A link to Solana AI's confidential chat for the models it serves. This model is not listed as served there.
Fine-tune it
Fine-tuning a few base models from this page and publishing the result here.
Creators and earnings
Pages for people who publish fine-tunes and adapters of this model, and what they earn.
Fund and bounties
Bounties for new open models, paid from a public fund. How the fund will work.
Contribute compute
How to give compute to open model work, with Akash.

Model data from the Hugging Face Hub 19 Nov 2026, 13:43 UTC. The repo. Found a mistake on this page? Tell us.