Gate News message. On April 10, Musk recently posted a reply on the X platform that sparked intense public speculation about the parameter size of Anthropic’s flagship model. In responding to a user’s follow-up question about the number of parameters in Grok 4.2, Musk confirmed: “0.5 trillion total parameters. Current Grok is half of Sonnet, and one-tenth of Opus. For its size, this is a very powerful model.” If we reverse-calculate based on Musk’s statement that “Grok 4.2 is one-tenth of Opus,” then the Claude Opus parameter count would be about 5 trillion, and Claude Sonnet about 1 trillion. It’s worth noting that Anthropic has never publicly disclosed the parameter size of any of its models. The figures above are only inferred from Musk’s remarks, not official data. Meanwhile, Musk revealed that SpaceX AI’s Colossus 2 supercomputing cluster is currently training 7 models in parallel, with the largest scale reaching 10 trillion parameters, and added: “There’s some catching up to do.” If the speculation holds true, Claude Opus with 5 trillion parameters would rank at the very top among currently known deployed models; and the 10-trillion-parameter model that xAI is training in parallel would become an important variable in the next round of competition.
Disclaimer: The information on this page may come from third parties and does not represent the views or opinions of Gate. The content displayed on this page is for reference only and does not constitute any financial, investment, or legal advice. Gate does not guarantee the accuracy or completeness of the information and shall not be liable for any losses arising from the use of this information. Virtual asset investments carry high risks and are subject to significant price volatility. You may lose all of your invested principal. Please fully understand the relevant risks and make prudent decisions based on your own financial situation and risk tolerance. For details, please refer to
Disclaimer.
Related Articles
OpenClaw Releases v2026.4.20 with Kimi K2.6 as Default Model
Gate News message, April 21 — OpenClaw, an open-source AI Agent platform, released v2026.4.20 on April 20, introducing significant updates to default model switching, plugin contract handling, and gateway storage optimization.
The release switches the default model for Moonshot's built-in
GateNews8m ago
OpenAI Launches ChatGPT Images 2.0 Model with Enhanced Visual Task Capabilities
Gate News message, April 21 — OpenAI unveiled ChatGPT Images 2.0 on April 22, significantly enhancing complex visual task processing capabilities. The model improves instruction understanding, object placement and relationship expression, and high-density
GateNews17m ago
SpaceX Agrees to Acquire AI Startup Cursor for Over $50 Billion
Gate News message, April 21 — According to The New York Times, SpaceX has agreed to acquire Cursor, a fast-growing artificial intelligence startup focused on code-writing software, for over $50 billion. The deal comes as SpaceX prepares for what could be one of the largest initial public offerings i
GateNews24m ago
Unauthorized Users Access Anthropic's Mythos AI Model, Company Confirms Investigation
Gate News message, April 22 — A small group of unauthorized users accessed Anthropic's new Mythos AI model on the same day the company announced plans to release it to select organizations for testing, Bloomberg News reported. The users, part of a private online forum, gained access to Mythos and
GateNews34m ago
Silicon Valley’s ‘monitoring the situation’ MTS meme becomes a 24/7 news machine delivered by a16z
a16z is backing "Monitoring the Situation," a 24/7 X livestream born from Polymarket meme culture, as tech VCs build their own news-industrial complex.
Summary
Andreessen Horowitz has helped launch "Monitoring the Situation" (MTS), a 24/7 livestream show on X, leaning into crypto-prediction mar
Cryptonews3h ago
Google launches Deep Research Max: supports MCP and can access enterprise private data
According to a Google DeepMind official blog announcement, Google launched a new generation of autonomous research agents, Deep Research and Deep Research Max, on April 21, 2026. They are built on Gemini 3.1 Pro, as the official release following the preview version provided through the Interactions API in December 2025. Both agents are now available in the paid tiers of the Gemini API in the form of a public preview, and new startups and enterprise users on Google Cloud will be able to connect gradually.
The two variants are positioned differently: interactive vs asynchronous depth
Google categorizes the two agents by use case: Deep Research
ChainNewsAbmedia5h ago
