MiniMax-Text-01

MiniMaxAI Not declared out 21 months ago

456B parameters (2 of 32 experts per token), text in, text out, 9.8M tokens of context. 1 provider serves it from $0.20 per million input tokens.

Verified 2026-10-25 Rank #641 in LLMs · score 73.2 method v1 Sources 4 Downloads 30d 2.8k · #571 of 619 tracked

Can I run it?274GB at INT4Estimateneeds a multi-GPU nodemethod

What does it cost?$0.20/1M incheapest of 1Minimax2026-10-24

Can I use it?Not declaredNo license declared on the model cardlicence file2026-10-24

Do people use it?2.8kdownloads, 30 days#571 of 619 we track · below the median+0 likes, downloads +6% this week657 likesHugging Face2026-10-24

Worth reading

Picked by rule from 4 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Most discussed · 19 HN pts

MiniMax-01, Advanced Text and Vision-Language Models

MiniMax-AI releases MiniMax-Text-01 and MiniMax-VL-01, large language and vision-language models using Linear Attention.

MiniMax-Text-01 positions a softmax attention layer after every 7 lightning attention layers.
The model employs a Top-2 routing strategy to select experts within its Mixture-of-Experts architecture.

github.com 2025-01-14 19 HN pts · 1 comments 2.4k words about the MiniMax 01 family family

Latest · 21 months ago

MiniMax-01: Scaling Foundation Models with Lightning Attention

MiniMax-01 releases MiniMax-Text-01 and MiniMax-VL-01 using lightning attention and a MoE architecture with 32 experts and 456 billion total parameters, where 45.9 billion are activated per token.

arxiv.org 2025-01-15 3 HN pts · 1 comments 365 words about the MiniMax 01 family family

4 best of 4 tracked

The 2 picks above, plus these 2. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

ArticleMiniMax-01 Is Now Open-Source: Scaling Lightning Attention for the AI Agent Eraminimaxi.com · 2025-01-15 · 1 HN pts · official
ArticleHow Minimax-01 Achieves 1M Token Context Length with Linear Attention (MIT)yacinemahdid.com · 2025-04-01 · 2 HN pts

Where to run it

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

#	Cheapest live row	In /1M	Out /1M	Context	Quant	Seen
1	Minimax	$0.20	$1.10	977K	unknown	Oct 24

All 1 provider →

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 1 API provider; 1 quantized build for local use.

As a service1Minimax
On own hardware1quantized builds
Built on11 fine-tunes · 0 adapters · 0 merges
In apps9Hugging Face Spaces

Top Spaces running it: qdqd/story-ai-2.0 ♥ 18 · TeacherPuffy/CreateBook ♥ 1

Will it fit?

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

Needs an 8×80 GB node even at INT4

About 274 GB at INT4, 547 GB at FP8, 1.09 TB 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

The story so far

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

Most coverage came in the first two weeks

Teal is the weights (2025-01-12), amber triangles are 4 dated articles (each links to it), grey ticks are repo commits.

What people built on it

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

2 derivatives on the Hub, led by 1 quantized

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

INT4 OPEA/MiniMax-Text-01-int4-sym-inc-preview for vLLM, SGLang 253 downloads

All variants →

Reference

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

Use-case evidence

3 use cases

What it is good at 3 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.

Chat and follow instructionsstructuraldecision-model estimate p=0.91 tag: conversational chat template present
Call tools and APIsstructuralchat template: tools
Read long documentsstructuralconfig: 10,240,000 context

Facts

21 fields

Listed in LLMs · MoE

Facts

MiniMax-Text-01 has 456B parameters. (safetensors metadata, observed 2026-10-24)
MiniMax-Text-01 is a mixture-of-experts model with 32 experts, routing 2 per token. (config.json, observed 2026-10-24)
Context length is 10,240,000 tokens (9.8M). (config.json max_position_embeddings, observed 2026-10-24)
License: Not declared (friction: review). No license declared on the model card. (license file, observed 2026-10-24)
Created 2025-01-12, last modified 2025-07-03. (HF API, observed 2026-10-24)

Pipeline tag	text-generation
Architectures	MiniMaxText01ForCausalLM
Model type	minimax_text_01
Params by dtype	BF16: 454,844,733,440 · F32: 1,244,921,856
Size band	XXL
MoE	yes
Experts	32
Experts per token	2
Context length	10,240,000 (config.json max_position_embeddings)
Papers	arXiv:2501.08313
License	Not declared
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: not in registry · MiniMaxText01ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: not in registry · 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-Text-01: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	1.09 TB · Cluster (≥640 GB)
FP8 (1 byte/param)	547 GB · Multi-GPU node (<640 GB)
INT4 (0.5 bytes/param)	274 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

2 variants

Lineage

Original model Family: MiniMaxAI/MiniMax-01 2 derivatives on the Hub

finetune: 1quantized: 1

Top variants (2 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
wangzhang/MiniMax-M2.5-abliterated	finetune	—	229B	XXL	other	wangzhang	10	23
OPEA/MiniMax-Text-01-int4-sym-inc-preview	quantized	int4	63.2B	XL	—	OPEA	1	253

Repository changelog

25 commits

Repository changelog 25 commits

Date	Commit	SHA
2025-07-03	update README	a7351bf
2025-06-28	Upload config.json (#40)	293f84c
2025-06-27	Make model config compatible with Hugging Face MiniMax implementation (#39)	130e821
2025-06-16	Update README.md	a59aa9c
2025-05-13	Correct the spelling error in the ReadMe file: change "guild" to "guide" (#34)	c07865b
2025-05-12	Guide document for function calling (#32)	7466de1
2025-04-17	Update README.md Add vLLM deployment info. (#31)	90d900a
2025-04-10	Add MiniMax MCP repo link to README.md	15443c6
2025-03-18	Update README.md (#28)	de45dcb
2025-02-24	Update README.md (#26)	194fe1a
2025-02-22	Update README.md	83fffd2
2025-02-22	Update README.md	38a03e4
2025-02-20	Update README.md	b1fdcf3
2025-02-20	Upload wechat-qrcode.jpeg	fb6756d
2025-02-20	Update LICENSE-CODE	f9e2ede
2025-02-20	Create MODEL-CODE	8cc077e
2025-02-20	Update LICENSE-MODEL	17f4e48
2025-02-20	Update MiniMax Chat website link	8e947cd
2025-01-17	Fix world_size in QuickStart Code	4ca1b78
2025-01-15	Remove import that no longer exists in Transformers (#5)	372fb1d
2025-01-15	Add Transformers as the library (#4)	40a455a
2025-01-15	Update Model Card: Citation and Quickstart	7119fdf
2025-01-14	MiniMax Logo	283e097
2025-01-14	Initial Commit	e1640d7
2025-01-12	initial commit	45fa731

All providers

1 provider · 1 endpoint

Where to run 1 providers · 1 endpoints

Hosted endpoints observed for MiniMax-Text-01, 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: $0.20/M.

Provider	Model ref	Context	Quant	$/M in	$/M out	tok/s	TTFT	Tools	Status	Observed
Minimax openrouter	minimax/minimax-01	977K	unknown	$0.20	$1.10	—	—	—	live	2026-10-24

Articles and coverage

4 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-01 Is Now Open-Source: Scaling Lightning Attention for the AI Agent Era

minimaxi.com2025-01-15officialabout the MiniMax 01 familyHN 1 pts · 0 comments

Analysis3

MiniMax-01, Advanced Text and Vision-Language Models

github.com2025-01-14authorityabout the MiniMax 01 familyHN 19 pts · 1 comments2,423 words

MiniMax-AI releases MiniMax-Text-01 and MiniMax-VL-01, large language and vision-language models using Linear Attention. The text model has 456 billion total parameters with 45.9 billion activated per token and supports a 1 million token training context length extending to 4 million tokens during inference. The vision model uses a 303-million-parameter Vision Transformer and recommends int8 quantization for deployment.written, reviewed, cited

MiniMax-Text-01 positions a softmax attention layer after every 7 lightning attention layers.
The model employs a Top-2 routing strategy to select experts within its Mixture-of-Experts architecture.
Deployment requires converting specific MoE gate coefficients and heads to int8 while protecting the lm_head and embed_tokens.

MiniMax-01: Scaling Foundation Models with Lightning Attention

arxiv.org2025-01-15authorityabout the MiniMax 01 familyHN 3 pts · 1 comments365 words

MiniMax-01 releases MiniMax-Text-01 and MiniMax-VL-01 using lightning attention and a MoE architecture with 32 experts and 456 billion total parameters, where 45.9 billion are activated per token. MiniMax-Text-01 trains on contexts up to 1 million tokens and extrapolates to 4 million tokens during inference, while MiniMax-VL-01 utilizes 512 billion vision-language tokens. Benchmarks show these models match GPT-4o and Claude-3.5-Sonnet performance with 20-32 times longer context windows.written, reviewed, cited

How Minimax-01 Achieves 1M Token Context Length with Linear Attention (MIT)

yacinemahdid.com2025-04-01about the MiniMax 01 familyHN 2 pts · 0 comments1,151 words

Minimax-01 is an MIT-licensed, 460 billion parameter model that integrates Lightning Attention with a Mixture of Experts to support context windows up to 4 million tokens. The architecture hybridizes attention mechanisms using an 8:1 ratio of linear to softmax blocks, enabling efficient scaling while matching closed-source model performance. This approach allows the system to extrapolate inference capabilities significantly beyond its training window of 1 million tokens.written, reviewed, cited

Linear attention models struggle to match softmax performance in tasks requiring complex reasoning or long-range dependencies.
Lightning Attention minimizes cumsum operations by adopting a tiling strategy similar to FlashAttention to overcome training inefficiency.
The model requires a specific recipe detailed in sections 4 and 5 of the paper to achieve its reported scaling and performance.

Papers

1 paper

Papers 1

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).

MiniMax-01: Scaling Foundation Models with Lightning Attention this model

MiniMax, Aonian Li, Bangwei Gong +87 · 2025-01-14 · PDF · HF paper page (304 upvotes)

Similar models

3 models

Similar models

Qwen3-Coder-480B-A35B-Instruct · 480B
same size band (XXL), also MoE, same task (text-generation), more providers, permissive license
MiniMax-M1-80k · 456B
same size band (XXL), also MoE, same task (text-generation), permissive license
GLM-4.7 · 358B
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 21:41 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:13 UTC · Last verified 2026-10-25

Follow MiniMax-Text-01

Watch this model

/feeds/models/MiniMaxAI--MiniMax-Text-01.xml(RSS 2.0) · every change to MiniMax-Text-01
/feeds/families/MiniMaxAI--MiniMax-01.xml(RSS 2.0) · every change across the MiniMax-01 family

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