MiniMax-M2

MiniMaxAI MiniMax (Modified MIT) (modified-mit) out 12 months ago

229B parameters (8 of 256 experts per token), text in, text out, 192K tokens of context. 3 providers serve it from $0.26 per million input tokens.

Verified 2026-10-25 Rank #101 in LLMs · score 99.1 method v1 Sources 4 Downloads 30d 117k · #302 of 619 tracked

Can I run it?137GB at INT4Estimatefits 2×80 GBmethod

What does it cost?$0.26/1M incheapest of 2Minimaxfp82026-10-24

Can I use it?MiniMax (Modified MIT) (modified-mit)MIT with added terms; read the repo LICENSE before commercial hostinglicence file2026-10-24

Does it work?61on SWE-bench Verified#9 of 22 measuredleaderboard2025-11-24

Do people use it?117kdownloads, 30 days#302 of 619 we track · top 49%+4 likes, downloads −35% this week1,513 likesHugging Face2026-10-24

Worth reading

Picked by rule from 16 tracked pieces: the lab's own post, the most shared, a field report, the latest.

The lab's own post

MiniMax M2: open model for agents and code

minimaxi.com 2025-10-28 4 HN pts 142 words sells inference

Most discussed · 84 HN pts

MiniMax M2.7 Is Now Open Source

MiniMax released M2.7, an agentic model that improved its own code over 100 rounds without human direction, achieving a 30% performance boost.

M2.7 scores 76.5 on SWE Multilingual and 52.7 on Multi SWE Bench, testing closer to real world engineering scenarios.
The model maintains 97% skill compliance across 40 complex skills on MM Claw, each skill exceeding 2,000 tokens.

firethering.com 2026-04-12 84 HN pts · 36 comments 1.1k words

Latest · 5 months ago

MiniMax M2 Technical Report Notes | Sebastian Raschka, PhD

MiniMax M2 is a 62-layer decoder-only transformer with 229.9 billion total parameters and a 196,608 token context window.

The full-attention model scored 90 on a 128K RULER task compared with 72 for the sliding-window version.
The fine-grained MoE version splits expert capacity into 128 smaller experts and selects eight per token.

sebastianraschka.com 2026-05-27

16 best of 16 tracked

The 5 picks above, plus these 11. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

PaperThe MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligencearxiv.org · undated
ArticleMiniMax-M3 vs MiniMax-M2: Model Comparisonartificialanalysis.ai · undated
ArticleComparing GPT-5.4, Opus 4.6, GLM-5.1, Kimi K2.5, MiMo V2 Pro and MiniMax M2.7codejam.info · 2026-04-18 · 6 HN pts
ArticleMiniMax-M2: Intelligence, Performance and Price Analysisartificialanalysis.ai · 2025-10-25 · 3 HN pts
CodeMiniMax-M2github.com · 2025-10-27 · 2 HN pts
ArticleMiniMax M2 | nextomoronextomoro.com · 2026-05-15
ArticleGitHubminimax-m2.com · undated
ArticleMiniMax M2 Review 2025 Speed & Featuresskywork.ai · undated
ArticleMiniMax M2 Review: Benchmarks, Pricing, Setup, Free APIbinaryverseai.com · 2026-02-03
ArticleMiniMax-M2 - In-depth Analysiscanopywave.com · 2026-01-30
ArticleMiniMax-M2: Technical Deep Dive into Interleaved Thinking for Agentic Coding Workflows - MarkTechPostmarktechpost.com · 2025-12-01

Show all 11

Where to run it

Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.

2 providers from $0.26

2 providers with a live price, from $0.26 to $0.30 per million input tokens; the median is $0.28. One dot per provider, cheapest in teal.

#	Cheapest 3	In /1M	Out /1M	Context	Quant	Seen
1	Minimax	$0.26	$1.02	200K	fp8	Oct 24
2	Google	$0.30	$1.20	192K	unknown	Oct 24
3	Featherless	—	—	—	—	Oct 25

All 3 providers →

Where it's used

Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.

Served by 3 API providers; 39 quantized builds for local use.

As a service3Featherless, Google, Minimax
On own hardware39quantized builds
Built on1915 fine-tunes · 2 adapters · 2 merges
In apps100+Hugging Face Spaces

Top Spaces running it: akhaliq/anycoder ♥ 3,321 · hadadxyz/ai ♥ 819 · umint/ai ♥ 258 · hfmlsoc/Lit_Review_with_LMs ♥ 3

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Fits two 80 GB GPUs at INT4; too big for one

About 137 GB at INT4, 274 GB at FP8, 549 GB at BF16: weights only, so leave room for the KV cache. Every expert must be resident, so total parameters count.

Estimate params × bytes per param × 1.2 · how we estimate

How it scores

Third-party measurements only; author-reported numbers stay in Reference, marked as such.

#9 of 22 on SWE-bench Verified so far

61. Each tick is a model on this site with a measured SWE-bench Verified row; author-reported numbers stay in Reference. Percent resolved. agent: mini-SWE-agent

Source · leaderboard · 2025-11-24

The story so far

Weights, every dated article, and repo commits on one rail.

8 dated articles and 30 repo commits so far

Teal is the weights (2025-10-22), amber triangles are 8 dated articles (each links to it), grey ticks are repo commits.

What people built on it

58 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

58 derivatives on the Hub, most of them quantized

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

GGUF unsloth/MiniMax-M2-GGUF for llama.cpp, Ollama, LM Studio 157.1k downloads MLX lmstudio-community/MiniMax-M2-MLX-8bit for Apple silicon 191.9k downloads AWQ QuantTrio/MiniMax-M2-AWQ for vLLM, SGLang 1.3M downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

1 row

Benchmarks 1 row

measured third-party leaderboard row (1)author-reported number from the model card or eval filesNo confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.

Benchmark	Value	Setting / notes	Reported by	Date
measured SWE-bench Verified · Write code with agents	61 percent resolved	agent: mini-SWE-agent	leaderboard	2025-11-24

Use-case evidence

5 use cases

What it is good at 5 use cases

Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.

Write code with agentsmeasuredSWE-bench Verified 61.0
Chat and follow instructionsstructuraldecision-model estimate p=0.94 tag: conversational chat template present
Call tools and APIsstructuralchat template: tool_calls
Reason through hard problemsstructuralchat template: <think>
Read long documentsstructuralconfig: 196,608 context

Facts

22 fields

Listed in LLMs · MoE

Facts

MiniMax-M2 has 229B parameters. (safetensors metadata, observed 2026-10-24)
MiniMax-M2 is a mixture-of-experts model with 256 experts, routing 8 per token. (config.json, observed 2026-10-24)
Context length is 196,608 tokens (192K). (config.json max_position_embeddings, observed 2026-10-24)
License: MiniMax (Modified MIT) (modified-mit) (friction: review). MIT with added terms; read the repo LICENSE before commercial hosting. (license file, observed 2026-10-24)
Created 2025-10-22, last modified 2025-12-23. (HF API, observed 2026-10-24)

Pipeline tag	text-generation
Library	transformers
Architectures	MiniMaxM2ForCausalLM
Model type	minimax_m2
Params by dtype	BF16: 1,230,021,632 · F32: 48,774,656 · F8_E4M3: 227,410,968,576
Size band	XXL
MoE	yes
Experts	256
Experts per token	8
Context length	196,608 (config.json max_position_embeddings)
Papers	arXiv:2504.07164, arXiv:2509.06501, arXiv:2509.13160
License	MiniMax (Modified MIT) (modified-mit) (other)
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · MiniMaxM2ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · MiniMaxM2ForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed

vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for MiniMax-M2: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	549 GB · Multi-GPU node (<640 GB)
FP8 (1 byte/param)	274 GB · Multi-GPU node (<640 GB)
INT4 (0.5 bytes/param)	137 GB · Multi-GPU node (<640 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra. MoE: all experts must be resident, so total parameters count.

All variants

40 variants

Lineage

Original release (shipped quantized by the lab) Family: MiniMaxAI/MiniMax-M2 58 derivatives on the Hub

quantized: 39finetune: 15adapter: 2merge: 2

Top variants (40 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
QuantTrio/MiniMax-M2-AWQ	quantized	AWQ	229B	XXL	apache-2.0	QuantTrio	10	1.3M
lmstudio-community/MiniMax-M2-MLX-8bit	quantized	MLX	229B	XXL	mit	lmstudio-community	1	191.9k
lmstudio-community/MiniMax-M2-MLX-4bit	quantized	MLX	229B	XXL	mit	lmstudio-community	0	176k
lmstudio-community/MiniMax-M2-MLX-6bit	quantized	MLX	229B	XXL	mit	lmstudio-community	0	172.6k
unsloth/MiniMax-M2-GGUF	quantized	GGUF	—	—	mit	unsloth	90	157.1k
bartowski/MiniMaxAI_MiniMax-M2-GGUF	quantized	GGUF	—	—	—	bartowski	11	22.4k
lemuralabs/MiniMax-M2-Pruned-25	finetune	—	173B	XXL	mit	lemuralabs	36	20.8k
mlx-community/MiniMax-M2-4bit	quantized	-4bit	229B	XXL	mit	mlx-community	10	18.3k
lmstudio-community/MiniMax-M2-GGUF	quantized	GGUF	—	—	mit	lmstudio-community	1	17.3k
redponike/MiniMax-M2-GGUF	quantized	GGUF	—	—	mit	redponike	2	15.6k
cturan/MiniMax-M2-GGUF	quantized	GGUF	—	—	mit	cturan	15	12.9k
DevQuasar/MiniMaxAI.MiniMax-M2-GGUF	quantized	GGUF	—	—	—	DevQuasar	15	11k
cerebras/MiniMax-M2-REAP-162B-A10B	quantized	—	162B	XXL	other	cerebras	80	6.4k
OsaurusAI/MiniMax-M2.7-Small-JANGTQ	finetune	—	12.5B	L	other	OsaurusAI	26	7.6k
ModelCloud/MiniMax-M2-GPTQMODEL-W4A16	quantized	GPTQ	229B	XXL	other	ModelCloud	3	8.5k
bullerwins/MiniMax-M2-GGUF	quantized	GGUF	—	—	mit	bullerwins	6	7.7k
tiny-random/minimax-m2	finetune	—	3.31M	S	—	tiny-random	0	6.9k
mlx-community/MiniMax-M2-8bit	quantized	-8bit	229B	XXL	mit	mlx-community	8	4.7k
mlx-community/MiniMax-M2-3bit	quantized	—	229B	XXL	mit	mlx-community	3	4.4k
cerebras/MiniMax-M2-REAP-139B-A10B	quantized	—	139B	XXL	other	cerebras	18	3.2k
lukealonso/MiniMax-M2-NVFP4	quantized	NVFP4	115B	XXL	—	lukealonso	14	3.2k
cerebras/MiniMax-M2-REAP-172B-A10B	quantized	—	173B	XXL	other	cerebras	23	2.5k
Intel/MiniMax-M2-gguf-q2ks-mixed-AutoRound	quantized	gguf	—	—	—	Intel	8	2.8k
moxin-org/MiniMax-M2-GGUF	quantized	GGUF	—	—	mit	moxin-org	6	2.8k
cyankiwi/MiniMax-M2-AWQ-4bit	quantized	AWQ	36.8B	XL	other	cyankiwi	4	2.5k
lemuralabs/MiniMax-M2-Pruned-55	finetune	—	106B	XXL	mit	lemuralabs	5	2.3k
mradermacher/MiniMax-M2-i1-GGUF	quantized	GGUF	—	—	other	mradermacher	8	2k
mlx-community/MiniMax-M2-5bit	quantized	—	229B	XXL	mit	mlx-community	2	2.2k
mlx-community/MiniMax-M2-6bit	quantized	—	229B	XXL	mit	mlx-community	2	2.2k
ilintar/MiniMax-M2-GGUF	quantized	GGUF	—	—	mit	ilintar	3	1.8k
ModelCloud/MiniMax-M2-BF16	finetune	-BF16	229B	XXL	mit	ModelCloud	3	1.6k
yujiepan/minimax-m2-tiny-random	finetune	—	3.31M	S	—	yujiepan	0	1.4k
anikifoss/MiniMax-M2-HQ4_K	quantized	GGUF	—	—	mit	anikifoss	2	1.1k
turboderp/MiniMax-M2-exl3	quantized	exl3	—	—	mit	turboderp	8	213
ArtusDev/MiniMaxAI_MiniMax-M2-EXL3	quantized	EXL3	—	—	mit	ArtusDev	5	355
a2s-ai/MiniMax-M2-AWQ	quantized	AWQ	229B	XXL	apache-2.0	a2s-ai	0	562
KikoCis/minimax-m2-expert-prune-heal	quantized	GGUF	—	—	apache-2.0	KikoCis	1	439
gauravmm/MiniMax-M2.5-REAP-139B-A10B-NVFP4	quantized	NVFP4	78.8B	XL	other	gauravmm	1	345
avtc/MiniMax-M2-GPTQMODEL-W4A16	quantized	GPTQ	229B	XXL	other	avtc	0	388
morriszjm/MiniMax-M2.5-192e	quantized	—	173B	XXL	other	morriszjm	0	277

Repository changelog

30 commits

Repository changelog 30 commits

Date	Commit	SHA
2025-12-23	Fix README.md typo (#50)	757303d
2025-12-08	improve model identity in chat template	f62f643
2025-11-13	Update modeling_minimax_m2.py	56992d0
2025-11-13	Update LICENSE (#39)	c1be0dc
2025-11-07	Fix chat_template (#32)	9354a5c
2025-11-06	update README	de27025
2025-11-06	Update docs/transformers_deploy_guide.md	eb390b8
2025-11-06	Prepare support transformers (#24)	78ba357
2025-10-29	update guides	c2b7e11
2025-10-29	update wechat qrcode	7584a2e
2025-10-29	update guide	3af90e1
2025-10-29	add wechat on README	d1b3a66
2025-10-29	upload wechat QRCode	1b75594
2025-10-29	Improve docs (#23)	197d677
2025-10-29	update mlx guide	db1917a
2025-10-29	update mlx-deploy-guide	cad818a
2025-10-29	Add MLX example (#22)	c706fa1
2025-10-28	add community showcases	457945b
2025-10-27	fix config typo	44cefa6
2025-10-27	update README	3f46446
2025-10-27	Adding `transformers` as the library name (#10)	2c35de5
2025-10-27	Update README.md	14ec561
2025-10-27	update bench figure	59f758d
2025-10-27	update README	1fe6a45
2025-10-27	update guides	a970723
2025-10-27	update sglang guide	4167c37
2025-10-27	Update generation_config.json	e977417
2025-10-27	update top_k	48a6a46
2025-10-27	update README	00b05bd
2025-10-27	Update README.md (#6)	c858585

All providers

3 providers · 3 endpoints

Where to run 3 providers · 3 endpoints

Hosted endpoints observed for MiniMax-M2, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.26/M.

Provider	Model ref	Context	Quant	$/M in	$/M out	tok/s	TTFT	Tools	Status	Observed
Minimax openrouter	minimax/minimax-m2	200K	fp8	$0.26	$1.02	—	—	yes	live	2026-10-24
Google openrouter	minimax/minimax-m2	192K	unknown	$0.30	$1.20	—	—	yes	live	2026-10-24
Featherless hf	MiniMaxAI/MiniMax-M2	—	—	—	—	—	—	—	live	2026-10-25

Articles and coverage

16 pieces

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.

Official1

MiniMax M2: open model for agents and code

minimaxi.com2025-10-28officialHN 4 pts · 0 comments142 words

Analysis9

MiniMax M2.7 Is Now Open Source

firethering.com2026-04-12HN 84 pts · 36 comments1,101 words

MiniMax released M2.7, an agentic model that improved its own code over 100 rounds without human direction, achieving a 30% performance boost. The model won 9 gold medals, 5 silver medals, and 1 bronze across 22 machine learning competitions runnable on a single A30 GPU. It scores 56.22% on SWE-Pro and reduces live production incident recovery time to under three minutes on multiple occasions. Commercial use requires prior written authorization from MiniMax, while non-commercial use remains free.written, reviewed, cited

M2.7 scores 76.5 on SWE Multilingual and 52.7 on Multi SWE Bench, testing closer to real world engineering scenarios.
The model maintains 97% skill compliance across 40 complex skills on MM Claw, each skill exceeding 2,000 tokens.

MiniMax M2 Technical Report Notes | Sebastian Raschka, PhD

sebastianraschka.com2026-05-27authority

MiniMax M2 is a 62-layer decoder-only transformer with 229.9 billion total parameters and a 196,608 token context window. The architecture uses 256 fine-grained experts per MoE layer, activating 9.8 billion parameters per token via sigmoid gating. MiniMax estimates its internal Model Iteration System handles 30 to 50 percent of its daily RL iteration workload.written, reviewed, cited

The full-attention model scored 90 on a 128K RULER task compared with 72 for the sliding-window version.
The fine-grained MoE version splits expert capacity into 128 smaller experts and selects eight per token.
The reinforcement-learning objective includes a task-completion-time reward that decreases as completion time rises.

The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence

arxiv.orgauthority13,874 words

MiniMax introduces the M2 series, a Mixture-of-Experts language model family designed for agentic workflows. The flagship M2 is a 62-layer decoder-only Transformer pre-trained on 29.2T tokens containing 229.9B total parameters with only 9.8B activated per token. The latest M2.7 checkpoint demonstrates early self-evolution by autonomously debugging training runs and modifying its own scaffold. On agentic coding, M2.7 reaches 56.2 on SWE-bench Pro, 76.5 on SWE-bench Multilingual, and 52.7 on Multi-SWE-bench.written, reviewed, cited

M2 uses sigmoid gating for expert routing, allowing multiple experts to activate simultaneously without a zero-sum constraint.
The model departs from hybrid attention mechanisms in favor of full multi-head attention across all layers.
Hybrid Sliding Window Attention variants showed degraded performance on retrieval and multi-hop reasoning tasks during pre-training.

MiniMax-M3 vs MiniMax-M2: Model Comparison

artificialanalysis.aiauthority785 words

MiniMax-M3 scores 29 on the Artificial Analysis Intelligence Index compared to MiniMax-M2's estimated 19. Released in June 2026 versus October 2025, the newer model costs $0.222 per 1M tokens against $0.39 for the older version. MiniMax-M3 features a 1000k token context window and a 1.26s time to first token, improving upon the 205k window and 1.41s latency of MiniMax-M2.written, reviewed, cited

MiniMax-M3 supports text, image, and video input modalities while MiniMax-M2 supports only text.
MiniMax-M3 costs $0.06 per 1M tokens for cache hits, a feature not listed for MiniMax-M2.
MiniMax-M3 has 428B total parameters and 23B active parameters compared to MiniMax-M2's 230B and 10B.

Comparing GPT-5.4, Opus 4.6, GLM-5.1, Kimi K2.5, MiMo V2 Pro and MiniMax M2.7

codejam.info2026-04-18HN 6 pts · 2 comments1,457 words

GPT-5.4 achieved an average cleanliness score of 9 while GLM-5.1 scored 8 in the technical approach evaluation. Kimi K2.5 and MiniMax M2.7 were consistently ranked at the bottom with average scores of 5.6 and 5.2 respectively.generated, cited

MiniMax-M2: Intelligence, Performance and Price Analysis

artificialanalysis.ai2025-10-25authorityHN 3 pts · 1 comments889 words

Aligning to What? Rethinking Agent Generalization in MiniMax M2

huggingface.co2025-12-26authorityHN 2 pts · 0 comments970 words

MiniMax-M2

github.com2025-10-27authorityHN 2 pts · 0 comments1,602 words

MiniMax M2.5 matches Claude Opus at 1/33rd the cost

news.reading.sh2026-02-12HN 3 pts · 0 comments

All coverage tracked for this model (16) — append-only, nothing is ever removed

Title	Source	Published	First seen	Signals
The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence	arxiv.org	—	2026-10-07	exa
MiniMax-M3 vs MiniMax-M2: Model Comparison	artificialanalysis.ai	—	2026-10-07	exa
GitHub	minimax-m2.com	—	2026-10-07	exa
MiniMax M2 Review 2025 Speed & Features	skywork.ai	—	2026-10-07	exa
MiniMax M2 Technical Report Notes \| Sebastian Raschka, PhD	sebastianraschka.com	2026-05-27	2026-10-07	exa
MiniMax M2 \| nextomoro	nextomoro.com	2026-05-15	2026-10-07	exa
Comparing GPT-5.4, Opus 4.6, GLM-5.1, Kimi K2.5, MiMo V2 Pro and MiniMax M2.7	codejam.info	2026-04-18	2026-10-07	hn
MiniMax M2.7 Is Now Open Source	firethering.com	2026-04-12	2026-10-07	hn
MiniMax M2.5 matches Claude Opus at 1/33rd the cost	news.reading.sh	2026-02-12	2026-10-07	hn
MiniMax M2 Review: Benchmarks, Pricing, Setup, Free API	binaryverseai.com	2026-02-03	2026-10-07	exa
MiniMax-M2 - In-depth Analysis	canopywave.com	2026-01-30	2026-10-07	exa
Aligning to What? Rethinking Agent Generalization in MiniMax M2	huggingface.co	2025-12-26	2026-10-07	hn
MiniMax-M2: Technical Deep Dive into Interleaved Thinking for Agentic Coding Workflows - MarkTechPost	marktechpost.com	2025-12-01	2026-10-07	exa
MiniMax M2: open model for agents and code	minimaxi.com official	2025-10-28	2026-10-07	hn
MiniMax-M2	github.com	2025-10-27	2026-10-07	hn
MiniMax-M2: Intelligence, Performance and Price Analysis	artificialanalysis.ai	2025-10-25	2026-10-07	hn

Papers

3 papers

Papers 3

arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).

WebExplorer: Explore and Evolve for Training Long-Horizon Web Agents cited in card

Junteng Liu, Yunji Li, Chi Zhang +12 · 2025-09-08 · PDF · HF paper page (83 upvotes)
FinSearchComp: Towards a Realistic, Expert-Level Evaluation of Financial Search and Reasoning cited in card

Liang Hu, Jianpeng Jiao, Jiashuo Liu +20 · 2025-09-16 · PDF · HF paper page (31 upvotes)
R2E-Gym: Procedural Environments and Hybrid Verifiers for Scaling Open-Weights SWE Agents cited in card

Naman Jain, Jaskirat Singh, Manish Shetty +3 · 2025-04-09 · PDF · HF paper page (2 upvotes)

Similar models

3 models

Similar models

MiniMax-M2.5 · 229B
same size band (XXL), also MoE, same task (text-generation), more providers
MiniMax-M2.7 · 229B
same size band (XXL), also MoE, same task (text-generation), more providers
Qwen3-235B-A22B · 235B
same size band (XXL), also MoE, same task (text-generation), more providers, permissive license

Sources

4 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:33 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:00 UTC · Last verified 2026-10-25

Follow MiniMax-M2

Watch this model

/feeds/models/MiniMaxAI--MiniMax-M2.xml(RSS 2.0) · every change to MiniMax-M2
/feeds/families/MiniMaxAI--MiniMax-M2.xml(RSS 2.0) · every change across the MiniMax-M2 family

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.