qwen

These two words—"open source"—may need to be reinterpreted in the future.

Reuters reports today that Alibaba is preparing to add a commercial clause to the next version of its open-source Qwen model. Large commercial customers would share revenue, and the rules are expected to be implemented as early as next week. How much would be shared is still being negotiated, and the percentage has not been set.

This approach is not Alibaba’s invention. Last month, Moonshot released Kimi K3. In its license agreement, there was a clause stating that companies that use the model to provide external services or have annual sales exceeding $20 million must return and sign a separate commercial agreement. Insiders say the maximum cut could be up to 30%.

Alibaba’s past pricing logic was simple: if the model was hosted on its own cloud, you paid. If customers moved it back to run in their own data centers, there was essentially no charge. Now, that line is being pushed upward.

Google $GOOGL and OpenAI have both been closed source, with weights not even provided for download. After this batch of open-source models from China has caught up in capability, their business models have started to tighten as well. The CEO of a U.S. cloud service provider confirmed that they have already signed a commercial agreement, saying it is a proven freemium model.

Model weights can still be downloaded, and the code can still be modified. The day it really earns money, the bill will show up on time.

#AI #开源大模型 #Qwen

🚨 $QWEN JUST FLIPPED THE AI COIN — THIS IS A LIQUIDITY SHIFT, NOT A BLIP 📉

Entry: 46% US token share 📊
Target: GPT-5.6 Sol premium squeeze 💥
Stop Loss: Independent benchmark validation ⚠️

✅ Alibaba's Qwen3.8-Max just threw a 2.4T-parameter haymaker at OpenAI's crown. The bid is stacking — Chinese-origin models now command up to 46% of US developer token volume, up from a 4.5% foothold six months ago. That's not a trendline, that's a takeover. 📈

💡 The real war is in the pricing trenches. OpenAI slashed its Luna tier 80% just three weeks post-launch — a textbook volume defense while the Sol flagship holds its premium. Meanwhile, Qwen's open-weight strategy flips the playbook: 95B active params per request and a 16-day autonomous coding demo tell the street that efficiency is the new edge. 🔍

🤔 When Chinese open models deliver 60-90% cost savings with a 2-point benchmark gap, how long before the routing algorithms abandon the premium entirely? 👇

⚠️ Not financial advice. Always manage your risk. 🛡️

🏷️ #QWEN #AIWar #OpenSource #TechTrade #CryptoAlpha

🎯 🦈

Alibaba just flexed hard with Qwen3.8 Max 👀

2.4 trillion parameters and they’re straight-up saying it’s only behind Anthropic’s Fable 5.

No independent benchmarks yet though, so take the ranking with a big grain of salt.

Still, open weights coming soon and the Chinese labs are clearly not slowing down after Kimi K3.

Interesting times. $BABA #AI #Qwen #ClaudeFable5 #rsshanto

$AI TOKENS HEATING UP AS QWEN3.8 DROPS WITH 24 TRILLION PARAMETERS 🔥

The Alibaba Q&A team just confirmed Qwen3.8 is landing soon and will be open-source. This is a 24-trillion parameter model — performance rivaling the top-tier AI systems, second only to Fable 5. The model is already live on Alibaba's Token Plan and Qoder platforms.

Massive AI compute demand means blockchain-based GPU and data networks could see fresh interest. Keep an eye on volume for tokens tied to decentralized inference.

Do you think this shifts attention back to AI narratives in crypto?

Not financial advice. Always manage your risk.

#AI #Qwen #Altcoins #CryptoNews

🔥

Alibaba Qianwen officially announced: the Qwen3.8 large model is set to be released and open-sourced, with parameter size reaching 2.4 trillion—one of the largest open-source models in China today.

Of note, the Qwen3.8-Max preview version has already been launched on Alibaba’s Token Plan, Qoder, and the QoderWork platform, allowing developers to try it out in advance.

With a 2.4T parameter scale, this generation of Qianwen is expected to see significant improvements in reasoning capability, multimodal fusion, and long-context processing. The open-sourcing strategy means developers can deploy it locally, which is an important variable for independent innovation in the AI application layer.

Competition in the AI space is shifting from the “parameter race” to the “application deployment race.” The open-sourcing of Qwen3.8 will accelerate this process.

#AI #Qwen

QWEN-AGENTWORLD IS REDEFINING HOW AGENTS LEARN AND INTERACT WITH REAL ENVIRONMENTS ⚡

The release of Qwen-AgentWorld marks a shift from general language models to models trained specifically for environment modeling. By leveraging over 10 million real-world interaction trajectories, this framework covers everything from terminal commands to complex Android GUI navigation.

The performance metrics are the real story here, with the 397B parameter model outperforming current industry leaders in simulation quality across seven major domains. This move toward world modeling suggests we are entering a new phase of agent capability where cross-domain transfer happens without needing constant fine-tuning.

Do you think world modeling is the final piece needed for autonomous agents to go mainstream?

Not financial advice. Always manage your risk.

#Qwen #AI #TechTrends #MachineLearning #Innovation

⚡

🍎 Apple + Alibaba = a new step in the AI race! 🤖
Apple has opened the door for integrating a Qwen model from Alibaba with Siri and writing tools on Mac devices in China.
🔹 More detailed responses
🔹 Image and document analysis
🔹 Text and image generation
🔹 Direct integration into the macOS experience
💻 The feature is available on macOS 26.6+ and requires enabling it and signing in to a Qwen account.
Most important?
This step reflects a growing trend toward partnerships between tech giants and local AI models to meet the needs of different markets.
Will we see similar partnerships from Apple in other markets?

#Apple #Alibaba #Qwen #AI #ArtificialIntelligence

🦈 $BABA 'S AI ARMY JUST GRABBED 46% OF US TOKEN FLOW — OPENAI IS BLEEDING SHARE 📈

📊 The institutional shift is unmistakable. Alibaba's Qwen3.8-Max — a 2.4 trillion parameter MoE beast with only 95B active per request — just landed in the US market and Chinese models have surged from 4.5% to 46% of developer token volume in under a year. 💡 That's not a blip; that's a structural takeover of the mid-tier compute layer.

⚡ OpenAI's response? An 80% price slash on their Luna tier just 3 weeks after launch — a textbook defensive move to protect volume share. The gap on Terminal-Bench is now just 2.2 points, and on OSWorld-Verified, Qwen actually leads. The open-weight wave is compressing the US frontier advantage to a 2-3 month window, per Berkeley's Ion Stoica.

💬 The real question: when open-weight models hit 90% of GPT-5.6's performance at a fraction of the cost, who pays for closed-source premiums? 👇

⚠️ Not financial advice. Always manage your risk. 🛡️

🏷️ #BABA #AI #Qwen #TechCompetition #Crypto

🔥 💎

Alibaba’s Qwen3.8-Max with 2.4T parameters gains traction as open weights neared; CrowdStrike reports AI-enabled attacks up 89% year-over-year. AI race and threats intensify. $BTC $ETH $SOL #AI #Alibaba #Qwen #CyberSecurity #Tech

Alibaba Cloud launches 'Peak-Valley Token': Qwen3.7 can be used at as low as 20% off during night hours

Alibaba Cloud's QoderWork has rolled out the 'Peak-Valley Token' pricing model, allowing users to automatically enjoy discounts from 10 PM to 8 AM the next day, with the Qwen3.7-Max model going as low as 20% off. Users can set scheduled tasks during the day for the Agent to execute autonomously at night.

Why it matters: Peak-Valley pricing lowers the barrier to entry for using AI models, paving the way for automated batch jobs with AI Agents to become the mainstream development paradigm.

#阿里云 #Qwen #AI #大模型 #cloudcomputing