Skip to content

Research lab

MiniMax

Chinese AI lab developing large language models and generative AI products, including the MiniMax model series used for text, voice, and video generation.

Known aliases

  • Hailuo AI
  • Minimax
  • MiniMaxAI
  • MiniMax AI
  • MiniMax-AI
  • MiniMax Code
  • MiniMax Group
  • MiniMax Group Inc.
  • Shanghai MiniMax Co., Ltd.

Current stories

leadershipOne report1 publisher

Character.AI pulled Disney characters after a September 18, 2025 cease-and-desist letter

Character.AI said on September 30, 2025 that it had removed Disney characters after a cease-and-desist letter dated September 18, 2025. It likens the user-made bots to fan fiction, so the takedown meets Disney's demand without settling the dispute.

Publishers:nbcnews.com

Reality

Evidence62
Adoption
Not scored
Hype gap0
Incentives55
Confidence65
buildOne report1 publisher

Count of 31 Chinese filing lists finds no separate entries for five labs' later general-purpose models

One researcher reading all 31 Chinese filing lists since 2023 finds no separate entries for five labs' later general-purpose models. That limits what the public can see of state review, but the researcher says it does not show regulators skipped it.

Publishers:lesswrong.com

Reality

Evidence55
Adoption
Not scored
Hype gap−5
Incentives
Not scored
Confidence55
productOne report1 publisher

Anthropic says 5,380 fake accounts let Moonshot pass Claude off as Kimi

Anthropic says Moonshot AI routed 300,000 Kimi customer requests to Claude over ten days through 5,380 fraudulent proxy accounts in Singapore and Japan. That averages about 56 requests per account, light enough that teams selling model access have to link accounts to one another to see it.

Reality

Evidence40
Adoption
Not scored
Hype gap+20
Incentives
Not scored
Confidence35
buildConfirmed3 publishers

Tech billionaires add a record $845bn in nine months on AI-driven US stocks

About 100 tech billionaires added a record $845bn in the first nine months of 2026, according to the Bloomberg Billionaires Index. The gains sit in a few US fortunes tied to AI share prices, so they describe this year's stock market better than next year's AI spending.

Perspective Coverage

5 publishers
Builder
Builder 18%
Operator
Operator 14%
Investor
Investor 68%

Reality

Evidence62
Adoption
Not scored
Hype gap+20
Incentives25
Confidence60
securityConfirmed8 publishers

NSA, CISA and FBI name six Chinese AI companies as industrial-scale distillers of US models

The conduct described is authorised API traffic at scale rather than an intrusion, the detection asked for is billing telemetry providers already hold, and on Beijing's role the advisory goes no further than likely government awareness.

Perspective Coverage

8 publishers
Builder
Builder 30%
Operator
Operator 50%
Investor
Investor 20%

Reality

Evidence55
Adoption
Not scored
Hype gap+20
Incentives55
Confidence60
buildOne report1 publisher

Anthropic traces nearly 300,000 Kimi requests to an Opus model through 5,380 fraudulent accounts

The allegation reaches Kimi's consumer output, and it arrives with two sets of numbers that do not fit inside each other. Anthropic has not dated the traffic, so nobody outside can tie it to a Kimi release.

Publishers:runtimewire.com

Reality

Evidence40
Adoption
Not scored
Hype gap+15
Incentives70
Confidence40
investConfirmed3 publishers

Anthropic's Alibaba distillation count rose fivefold while the accounts behind it shrank sevenfold

The September threat report puts 151 million Claude exchanges on 3,500 accounts Anthropic links to Alibaba, about 469 a day per account. The only figure denominated in money anywhere near it is a 2.7% share move.

Perspective Coverage

3 publishers
Builder
Builder 28%
Operator
Operator 35%
Investor
Investor 37%

Reality

Evidence45
Adoption
Not scored
Hype gap+20
Incentives65
Confidence50
buildConfirmed3 publishers

DeepSeek's new encoder-decoder splits inference into an 8B prefill and a 16B decode

V4.1-Flash retires the V4 Pro line and carries two active-parameter counts, 763B total with 8B on input tokens and 16B on output, so one sizing number no longer covers both phases of a request. Baseten had it running on day zero.

Perspective Coverage

3 publishers
Builder
Builder 40%
Operator
Operator 28%
Investor
Investor 32%

Reality

Evidence60
Adoption35
Hype gap+25
Incentives40
Confidence58
investConfirmed3 publishers

Zhipu fell four times as far as Tencent on Beijing's DeepSeek data probe

The Cyberspace Administration sent officials into DeepSeek and Moonshot AI to ask whether user requests were routed through Anthropic's Claude. Hong Kong investors sold the listed model developers hardest.

Perspective Coverage

3 publishers
Builder
Builder 27%
Operator
Operator 30%
Investor
Investor 43%

Reality

Evidence45
Adoption40
Hype gap+25
Incentives70
Confidence55
productOne report1 publisher

Chinese officials dismiss Amodei's proposed AI slowdown as fear-mongering

Anthropic's chief executive asked the industry to slow its most powerful models and said the pause needs China's cooperation. Beijing's foreign ministry called the idea fear-mongering three days before its own standards body warned about recursive self-improvement.

Publishers:restofworld.org

Reality

Evidence68
Adoption15
Hype gap+20
Incentives80
Confidence60
productOne report1 publisher

China's internet regulator questions DeepSeek and Moonshot staff over data Anthropic may have received

The Information reports that the Cyberspace Administration has questioned staff at both labs after Anthropic named seven Chinese companies for relaying customer requests through Claude. No penalty has been decided.

Publishers:thenextweb.com

Reality

Evidence52
Adoption42
Hype gap+8
Incentives72
Confidence55
leadershipOne report1 publisher

Anthropic accuses three rival labs of extracting 16 million Claude exchanges through fake accounts

Anthropic ties the traffic to DeepSeek, Moonshot and MiniMax through IP correlation, request metadata and unnamed industry partners. The DeepSeek campaign it quantifies is under 1 percent of the alleged total.

Publishers:anthropic.com

Reality

Evidence35
Adoption
Not scored
Hype gap+30
Incentives85
Confidence55
buildOne report1 publisher

Pooling per-job status polls into one Server Action cuts the client queue depth to one

Next.js dispatches Server Actions one at a time per client connection, so open-higgsfield collapses every in-flight generation poll into a single 4,000 ms tick and fans the status calls out on the server with Promise.all.

Publishers:dev.to

Reality

Evidence45
Adoption15
Hype gap+35
Incentives60
Confidence55
buildOne report1 publisher

Quantizing value residuals and casting softmax to FP8 speeds Nunchux's attention kernel 1.46x on an H200

The September 14th VC-Attention paper reports 1.46x on an H200 and 1.59x on a B200 against BF16 FlashAttention-4, measured at the attention call. How much of that reaches a finished clip depends on how much of your step is attention.

Publishers:runtimewire.com

Reality

Evidence55
Adoption18
Hype gap+18
Incentives70
Confidence58

Earlier coverage

  1. A 500 billion yuan pre-IPO target would price DeepSeek at 148 times reported revenue

    Invest · September 14, 2026 · Confirmed2 publishers

  2. NSA, CISA and FBI ask providers to quietly route suspected distillers to weaker models

    Build · September 10, 2026 · Confirmed2 publishers

  3. MiniMax's revenue nearly quadrupled. Its gross profit covers 7% of its adjusted loss.

    Invest · August 26, 2026 · Confirmed2 publishers

  4. The $100 sorority girl and the end of buying attention by the view

    Leadership · August 20, 2026 · One report1 publisher

  5. Mojo is Apache 2.0 now. The licence was never the hard part

    Build · August 19, 2026 · Confirmed3 publishers