📊 Full opportunity report: Second Only To Fable 5: Qwen3.8-Max Finally Shows Its Numbers — And The Claim Gets Complicated on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Alibaba’s Qwen3.8-Max has disclosed its benchmark results, confirming it as the second-leading model after Fable 5. The model features 2.4 trillion parameters and excels in multimodal tasks, with open weights expected next week.
Alibaba has officially released the detailed benchmark results for its Qwen3.8-Max model, confirming it as the second most powerful model globally after Fable 5. This development follows weeks of speculation and stealth previews, making it a significant milestone in AI model deployment and transparency.
On August 3, Alibaba published the full benchmark table for Qwen3.8-Max, revealing it has 2.4 trillion parameters and is built on the Qwen3.5 architecture with sparse mixture-of-experts. The model demonstrates top-tier performance on several benchmarks, including Terminal-Bench 2.1 (86.6), surpassing Claude models and only trailing GPT-5.6 Sol at 88.8. It also leads in PaperBench (93.0) and performs well in multimodal and agentic tasks, such as OSWorld-Verified (86.1) and Parametric CAD Bench (91.5).
Alibaba confirmed that open weights for Qwen3.8-Max will ship next week, alongside a smaller 27-billion-parameter checkpoint, Qwen3.8-27B, optimized for local deployment. The model’s active parameters are approximately 95 billion, with the total size at 2.4 trillion, indicating a model that requires multi-node datacenter infrastructure for hosting. The release of the full benchmark data marks a shift from stealth preview to transparency, with the model now broadly accessible via API and open weights expected soon.
For fifteen days the claim ran without a benchmark table. Today Alibaba published the table, the active-parameter count, and a weights timeline. The numbers are genuinely strong on the rows Alibaba chose — and twelve to fifteen points behind on the rows it didn’t.
▲ All performance figures: Alibaba’s own harnessThe claim shipped on a Sunday. The evidence shipped two weeks later. In between, the claim did its work.
“Second only to Fable 5” is true on the rows Alibaba chose and false on the rows it didn’t. Both halves below are from the same release.
“Qwen3.8 is going open-weight” describes three things with very different deployment realities.
OpenAI- and DashScope-compatible — a base-URL change to A/B against your current backend.
A multi-node datacenter artifact. At 95B active, no single machine serves it. A flag planted, not a deployment option.
The checkpoint that fits real hardware. Whether the agentic gains survive distillation is the question that decides whether next week matters.
Three Chinese frontier releases in seventeen days, each measured against the same export-controlled model. The contest is real; it is not the same thing as your workload.
- The generation jump is real and consistent across a dozen agentic rows, with a stated mechanism: RL-environment scaling.
- More disclosure than Kimi K3 shipped — full table, active-parameter count, weights timeline.
- If 2.4T lands under a permissive licence, the ceiling of “open weight” moves permanently.
- The 27B sibling could become the best local agent model on hardware people already own.
- Every number is Alibaba’s harness. Independent testing already tempered Kimi K3’s launch claims substantially.
- The paying use case still belongs to Fable 5 — twelve to fifteen points on deep software engineering.
- “Next week” comes from a company that sat on a finished benchmark table for fifteen days.
- Until the licence text exists, “going open-weight” is a press strategy, not a property of the model.
and it says “second only” depends entirely on which row you read.
Implications of Alibaba's Benchmark Disclosure
This announcement signifies a major step in AI transparency and competition. Alibaba’s disclosure of detailed benchmark results and open weights demonstrates a move toward greater openness in large language model deployment. The model’s high performance in key benchmarks and agentic capabilities suggests it could influence enterprise AI applications and challenge existing leaders like OpenAI and Anthropic. The release of open weights also expands access for researchers and developers, potentially accelerating innovation and adoption.
Local LLM Inference Optimization: A Comprehensive Guide to Quantization, Hardware Acceleration, and Efficient Private AI Deployment
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background on Alibaba’s AI Model Development
Over the past two weeks, Alibaba’s AI model activity was largely stealthy, with the model known as kaleb appearing on leaderboards and later confirmed as Qwen3.8-Max during the World AI Conference in Shanghai. The company had previously teased a 2.4 trillion-parameter model but withheld detailed benchmarks until now. The model's preview was accessible via a paid endpoint, sparking speculation about its capabilities and positioning in the market. The recent disclosure follows a pattern of Alibaba gradually revealing its progress, culminating in today’s comprehensive benchmark release.
"We are committed to open AI development and will ship open weights next week, enabling broader access and innovation."
— Alibaba spokesperson
As an affiliate, we earn on qualifying purchases.
Remaining Questions About Model Deployment and Licensing
While benchmark results are now confirmed, details about the licensing terms for the open weights remain unpublished. It is also unclear whether the 2.4 trillion-parameter model will be fully open-source or subject to restrictions, as previous Alibaba models have varied licenses. Additionally, the performance of the 27B checkpoint in practical, local deployment settings has yet to be demonstrated, and the long-term stability of agentic capabilities remains to be seen.
As an affiliate, we earn on qualifying purchases.
Next Steps for Alibaba’s Model Release and Adoption
Alibaba will release the open weights for Qwen3.8-Max next week, enabling researchers and developers to test and deploy the model locally. Monitoring how the model performs in real-world applications, especially in agentic tasks, will be key. Further benchmark results for the 27B checkpoint are expected, along with potential updates on licensing terms. Industry analysts will also watch for how competitors respond and whether Alibaba’s transparency influences market dynamics.
As an affiliate, we earn on qualifying purchases.
Key Questions
What are the key performance metrics of Qwen3.8-Max?
Qwen3.8-Max scores 86.6 on Terminal-Bench 2.1, 93.0 on PaperBench, and 86.1 on OSWorld-Verified, among other benchmarks, placing it second only to Fable 5 in overall performance.
Will the open weights be freely available for download?
Alibaba announced that open weights for Qwen3.8-Max will ship next week, but the licensing terms remain unpublished. It is expected that the weights will be accessible, though restrictions may apply depending on licensing decisions.
How does Qwen3.8-Max compare to other models in agentic tasks?
The model shows significant improvements in agentic benchmarks, with scores rising from 21.6 to 56.6 on DeepSWE and from 40.7 to 73.5 on FrontierSWE, indicating enhanced long-horizon reasoning and task execution capabilities.
What are the implications for AI research and deployment?
The detailed benchmark disclosure and upcoming open weights could accelerate AI research, democratize access, and challenge existing market leaders, fostering more competition and innovation in large language models.
Source: ThorstenMeyerAI.com
![Fil-C: Garbage In, Memory Safety Out [Video]](https://bestcadpapers.com/wp-content/uploads/2026/07/fil-c-garbage-in-memory-safety-out-video-featured-560x300.jpg)
