{"slug": "i-don-t-want-an-oligopoly-new-open-weight-ai-models-mount-a-comeback-against", "title": "\"I don't want an oligopoly\": New open-weight AI models mount a comeback against China", "summary": "Reflection introduced Beam, a 501-billion-parameter mixture-of-experts model with 23 billion active parameters, on Monday, and Mistral said Tuesday it is finishing Mistral Large 4, a 1-trillion-parameter multimodal model with 49 billion active parameters, as Western open-weight systems mount a comeback against Chinese models such as GLM-5.2 and Alibaba's Qwen 3.8-Max. Reflection plans to release Beam's weights later this month with tools for running, evaluating and fine-tuning it, while Mistral will initially offer Le Chonk through a moderated API and release its weights Oct. 27 after further reinforcement learning and safety testing. Reflection CEO Misha Laskin said \"once you're spending that amount on intelligence, you want to move from renting it to owning it yourself,\" while Mistral VP of science Pierre Stock said \"I don't want to live in the future in which any oligopoly controls closed access to this type of intelligence.", "body_md": "A new crop of powerful, Western open-weight AI systems are mounting a comeback, with fresh models from Reflection and Mistral narrowing China's recent lead.\nWhy it matters: These models and others will give businesses and governments alternatives to the closed ecosystems of OpenAI and Anthropic, as well as to Chinese options they may have eschewed for security reasons.\nBut putting powerful model weights into the world also makes their safeguards easier to remove, a concern that's growing as models become more capable in areas like cybersecurity.\nWhat they're saying: Both companies are arguing that raw intelligence isn't the only way to measure the AI race, and they present the notion of user control as a core opportunity and selling point.\n\"Once you're spending that amount on intelligence, you want to move from renting it to owning it yourself,\" Reflection CEO Misha Laskin told Axios. \"That's kind of where open source is very powerful because it is customizable at every level.\"\nMistral VP of science Pierre Stock said the themes he repeatedly hears from customers include control over data and intellectual property, business continuity, the ability to customize models and cost. \n\"I don't want to live in the future in which any oligopoly controls closed access to this type of intelligence,\" Stock said in an interview.\nZoom in: Reflection on Monday introduced Beam, a 501-billion-parameter mixture-of-experts model, with 23 billion parameters active at a time.\nReflection says Beam is competitive with China's GLM-5.2 and approaches Alibaba's Qwen 3.8-Max on some coding and agentic tasks. Its own published benchmarks show a mixed picture: Beam beats or roughly matches the Chinese models on some tests while trailing them on others.\nReflection says Beam can achieve reasoning performance comparable to GLM-5.2 with three to four times less computing firepower — an important consideration because the cost of running AI models is now as important to many customers as raw performance.\nReflection plans to release Beam's weights later this month, along with tools for running, evaluating and fine-tuning it.\nMeanwhile, Mistral said Tuesday it is finishing work on a new flagship model, Mistral Large 4, aka \"Le Chonk.\" It's a 1-trillion-parameter multimodal model with 49 billion active parameters, trained on 4,000 Nvidia Grace Blackwell GPUs over two months in Mistral's own European data centers.\nStock told Axios the company believes Le Chonk is the world's best open-weight model and said it can outperform closed models on certain tasks.\nBut Stock acknowledged that Mistral hasn't yet caught the leading closed models. \"We're not there yet on the frontier,\" he said.\nMistral is initially making the model available through a moderated API, with a version with fewer restrictions and broader cybersecurity capabilities being shared with select partners for testing. It plans to release the weights Oct. 27 after further reinforcement learning and safety testing.\nThe big picture: Chinese companies including Alibaba, Z.ai, Moonshot and DeepSeek have driven much of the recent momentum in open-weight AI.\nThe arrival of competitive models from Reflection and Mistral suggests Western labs may be narrowing a gap that had begun to look structural.\nYes, but: The same characteristic that gives customers control over an open-weight model also limits the developer's control over how it is used.\nOnce weights are publicly available, users can modify the model, including attempting to remove safeguards added by its creator.\nThat's particularly significant as models become more capable at cybersecurity and other potentially dangerous tasks.\nThe other side: Mistral and Reflection argue that openness can also improve safety.\nLaskin said having a larger ecosystem able to inspect models and hunt for vulnerabilities is safer than concentrating that work among a few hundred researchers inside closed labs. He said there should nevertheless be capability-based safety thresholds that determine whether any model — open or closed — should be released.\nStock acknowledged a trade-off in releasing weights but argued that making capable models available can accelerate cyber defenses and allow outside researchers to audit the technology. Mistral is using a staged release to give itself additional time for testing.\nOpen-weight models also ensure that those defending against cyberattacks have adequate tools. \"If we open weight as many models as possible, including ML4 — which is among the best models in the world on cyber — then you accelerate the defense part way more,\" Stock said.", "url": "https://wpnews.pro/news/i-don-t-want-an-oligopoly-new-open-weight-ai-models-mount-a-comeback-against", "canonical_source": "https://www.machinebrief.com/news/i-dont-want-an-oligopoly-new-open-weight-ai-models-mount-a-c-f45d", "published_at": "2026-10-06 13:05:06+00:00", "updated_at": "2026-10-06 14:17:12.194368+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-safety", "ai-startups", "ai-products"], "entities": ["Reflection", "Mistral", "Beam", "Mistral Large 4", "Misha Laskin", "Pierre Stock", "Alibaba", "Nvidia"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/i-don-t-want-an-oligopoly-new-open-weight-ai-models-mount-a-comeback-against", "markdown": "https://wpnews.pro/news/i-don-t-want-an-oligopoly-new-open-weight-ai-models-mount-a-comeback-against.md", "text": "https://wpnews.pro/news/i-don-t-want-an-oligopoly-new-open-weight-ai-models-mount-a-comeback-against.txt", "jsonld": "https://wpnews.pro/news/i-don-t-want-an-oligopoly-new-open-weight-ai-models-mount-a-comeback-against.jsonld"}}