• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Zephyr Predicts Non-Commercial Licenses for Chinese AI Models

    Zephyr outlines next steps for vendors including revenue-sharing and enterprise licensing deals.

    T(
    AD
    ZE
    9 Sources, 29d ago, first seen 29d ago

    TLDR

    Zephyr, a pseudonymous commentator focused on AI hardware and infrastructure, posted that Chinese open-weight model vendors will next release models under non-commercial licenses. The post describes inference revenue-sharing agreements and licensing for post-training. Enterprises could pay five million to ten million dollars per year for rights to post-train a vendor's models. These statements come from a social media thread on AI topics. The packet contains no first-party announcements or corroboration from other sources.

    Combined views

    161.8K

    9 Sources, first seen 29d ago

    Combined views

    161.8K

    9 Sources, first seen 29d ago

    1.5K likes
    1.5K likes
    92 comments
    219 saves
    48 reposts

    Sentiment

    Positive37.8%62.2%Negative

    Based on 37 sentiment-bearing replies from 37 accounts across 4 conversations.

    92 comments
    219 saves
    48 reposts
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    Sentiment

    Positive37.8%62.2%Negative

    Based on 37 sentiment-bearing replies from 37 accounts across 4 conversations.

    9 Sources

    @zephyr_z9That's why the next step for them will be releasing models under a non-commercial license Get an inference revenue-sharing agreement Licensing for post-training Chinese open-weight models is also coming An enterprise can pay $5M-$10M/year, which gives them the right to post-train the models of that vendor
    @DorialexanderI hope everyone claiming that EU AI sovereignty is just a cloud issue has provisioned for license fees.
    @teortaxesTexEuros won't get to thanklessly freeride on Chinese work (so long as they don't go rogue). At best they will have to post-train very strong models already released under permissive licenses. This gets them maybe a year. Fine, but after that, gotta get serious.
    @xlr8harderhonestly surprised i haven't seen any paranoia about glm 5.3's model license introducing the idea of them performing an unspecified security review clearly this is just a CCP foothold into western companies
    @adityaagIt's actually really hard to trust these frontier labs when they say "don't worry, we won't use your data for training". This is a huge reason to use OSS models. What a weird world. Use the 🇨🇳 models to make sure your stuff is private. Also why it is critical to have 🇺🇸 OSS.
    @SabrinaHalperSo much discussion about whether we should be using Chinese open-source models in the US. But how could they actually hurt us? Matan Grimberg @matanSF breaks down the biggest risks- 1. Intentional vulnerabilities. A model could contain backdoors, sleeper behavior, or other weaknesses deliberately introduced into the weights, potentially sitting dormant until the right conditions trigger them. 2. Unintentional vulnerabilities. A model built in China will naturally be trained and optimized around Chinese users, codebases, languages, and data. It could perform incredibly well overall while having subtle blind spots or security weaknesses in Western systems that nobody intentionally put there. The third we may not know for years. The most concerning failures might not show up in today’s benchmarks or security evaluations. If models become deeply embedded across companies and infrastructure, vulnerabilities could emerge only after we’ve become dependent on them.
    @matanSFRT @SabrinaHalper: So much discussion about whether we should be using Chinese open-source models in the US. But how could they actually hu…
    @thdxrin this whole USA vs China thing OpenAI and Anthropic aren't relevant because they're positioned differently them building better models doesn't hurt china at all the competitor has to be - american - open source - enough compute to do inference at scale that can shift things
    @Dan_Jeffries1The answer has always been and always will be, American open weights. One of these days a company will come roaring out of the gate in the States and figure out the open weights business model. In the Linux era, Red Hat defined the open source business model and others, like MongoDB and Hashicorp, refined it. Same will happen with AI. The likely answer is monitizing hardware (Fireworks/Together/Apple) and/or commoditizing today's software stacks into a single unified agentic interface and abstracting up the stack and selling proprietary software/memories/tools/skills on top of it. But it could be something else. Someone will find it. Let's get back to building the future here.

    9 Sources

    @zephyr_z9That's why the next step for them will be releasing models under a non-commercial license Get an inference revenue-sharing agreement Licensing for post-training Chinese open-weight models is also coming An enterprise can pay $5M-$10M/year, which gives them the right to post-train the models of that vendor
    @DorialexanderI hope everyone claiming that EU AI sovereignty is just a cloud issue has provisioned for license fees.
    @teortaxesTexEuros won't get to thanklessly freeride on Chinese work (so long as they don't go rogue). At best they will have to post-train very strong models already released under permissive licenses. This gets them maybe a year. Fine, but after that, gotta get serious.
    @xlr8harderhonestly surprised i haven't seen any paranoia about glm 5.3's model license introducing the idea of them performing an unspecified security review clearly this is just a CCP foothold into western companies
    @adityaagIt's actually really hard to trust these frontier labs when they say "don't worry, we won't use your data for training". This is a huge reason to use OSS models. What a weird world. Use the 🇨🇳 models to make sure your stuff is private. Also why it is critical to have 🇺🇸 OSS.
    @SabrinaHalperSo much discussion about whether we should be using Chinese open-source models in the US. But how could they actually hurt us? Matan Grimberg @matanSF breaks down the biggest risks- 1. Intentional vulnerabilities. A model could contain backdoors, sleeper behavior, or other weaknesses deliberately introduced into the weights, potentially sitting dormant until the right conditions trigger them. 2. Unintentional vulnerabilities. A model built in China will naturally be trained and optimized around Chinese users, codebases, languages, and data. It could perform incredibly well overall while having subtle blind spots or security weaknesses in Western systems that nobody intentionally put there. The third we may not know for years. The most concerning failures might not show up in today’s benchmarks or security evaluations. If models become deeply embedded across companies and infrastructure, vulnerabilities could emerge only after we’ve become dependent on them.
    @matanSFRT @SabrinaHalper: So much discussion about whether we should be using Chinese open-source models in the US. But how could they actually hu…
    @thdxrin this whole USA vs China thing OpenAI and Anthropic aren't relevant because they're positioned differently them building better models doesn't hurt china at all the competitor has to be - american - open source - enough compute to do inference at scale that can shift things
    @Dan_Jeffries1The answer has always been and always will be, American open weights. One of these days a company will come roaring out of the gate in the States and figure out the open weights business model. In the Linux era, Red Hat defined the open source business model and others, like MongoDB and Hashicorp, refined it. Same will happen with AI. The likely answer is monitizing hardware (Fireworks/Together/Apple) and/or commoditizing today's software stacks into a single unified agentic interface and abstracting up the stack and selling proprietary software/memories/tools/skills on top of it. But it could be something else. Someone will find it. Let's get back to building the future here.