📊 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 has released detailed benchmark results for its Qwen3.8-Max model, confirming it as the second most powerful model after Fable 5. The company also announced open weights will be available next week, with a smaller 27B version for local deployment.
Alibaba has officially published the benchmark results for its Qwen3.8-Max model, confirming it as the second most powerful large language model after Fable 5. The release includes detailed performance metrics across multiple benchmarks, and the company announced that open weights will be available next week, along with a smaller 27B parameter version tailored for local deployment. This development marks a significant step in Alibaba’s AI model offerings and clarifies its competitive standing in the industry.
On August 3, Alibaba made public the full benchmark table for its Qwen3.8-Max model, which was previously previewed in July under the codename ‘kaleb.’ The model features approximately 2.4 trillion total parameters, with an active parameter count of roughly 95 billion per query, utilizing a sparse mixture-of-experts architecture based on Qwen3.5. The benchmarks show that Qwen3.8-Max outperforms models like Claude Opus 4.8 and Claude Fable 5 on several tests, but remains behind GPT-5.6 Sol at maximum effort.
The model demonstrates strong performance in multimodal tasks — including text, image, and video inputs — and excels in agentic and long-horizon tasks, with significant improvements over its predecessor, DeepSWE. However, it trails notably on deep software-engineering benchmarks such as SWE-bench Pro and FrontierSWE, with gaps of 12-15 points compared to Fable 5. The company also confirmed that the open weights will be shipped next week, enabling broader access, although the full licensing details are still unpublished. The 2.4 trillion parameters are hosted in a multi-node data center environment, making self-hosting infeasible for most users.
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 Benchmark Results for Alibaba's AI Strategy
The publication of these benchmark scores confirms Alibaba's position as a leading player in large language models, with a model that rivals the best in the industry in several key areas. The detailed performance metrics and open weights signal a move toward greater transparency and accessibility in the AI ecosystem, potentially influencing industry standards and competitive dynamics. The improvements in agentic and long-horizon tasks suggest Alibaba's focus on practical, real-world applications, which could accelerate adoption and innovation in AI-driven solutions.
However, the model's limitations on software-engineering benchmarks highlight ongoing challenges in achieving broad-based excellence across all domains. The upcoming release of the open weights will be closely watched to assess how well the model can be adapted for diverse deployment scenarios, especially at the smaller, more manageable 27B size designed for local hardware. Overall, the development underscores Alibaba's strategic emphasis on both high performance and openness in AI development.
Mastering Small Language Models: A Practical Guide to Building Lightweight NLP Systems with Python, Transformers, and Quantization Techniques
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Recent Developments Leading to Qwen3.8-Max Benchmark Reveal
Over the past two weeks, Alibaba's AI models have been shrouded in secrecy, with the company teasing a model claimed to be 'second only to Fable 5.' The model was first previewed in July under the codename 'kaleb,' which was later identified as Qwen3.8-Max during the World AI Conference in Shanghai. Prior to today’s full disclosure, Alibaba released limited information, emphasizing a 2.4 trillion parameter count but withholding detailed benchmarks and licensing terms.
During this period, other models like Moonshot's Kimi K3 and the anonymous 'Claude' model gained attention, but Alibaba’s strategic release of benchmark data and specifications now clarifies its standing. The company’s approach has combined staged announcements with a focus on transparency, culminating in today’s comprehensive publication of performance metrics, which confirm the model’s capabilities and competitive position.
"We are committed to transparency and openness, and our upcoming open weights will enable broader access and innovation."
— Alibaba spokesperson
AI Systems Performance Engineering: Optimizing Model Training and Inference Workloads with GPUs, CUDA, and PyTorch
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Remaining Questions About Licensing and Deployment
While Alibaba has announced that open weights will be shipped next week, the exact licensing terms and usage restrictions remain unpublished, raising questions about accessibility and commercial use. Additionally, the full benchmark table for the 27B version has not yet been released, making it unclear how well the smaller model will perform in practical deployment scenarios. The long-term impact of the agentic improvements on real-world applications also remains to be seen.
LAFVIN ESP32-S3 1.69" LCD Development Board with Camera, AI Vision Voice Development Kit, Programmable IoT Board with Mic Speaker for STEM Education
- Powerful Microcontroller: ESP32-S3 with 16MB Flash and 8MB PSRAM
- AI Vision & Voice Capabilities: Camera and audio for AI interactions
- Supports OpenCV & YOLO: Face tracking and human pose estimation
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps: Open Weights Release and Model Deployment
Alibaba plans to release the open weights of Qwen3.8-Max next week, which will allow third-party developers and organizations to evaluate and adapt the model for their needs. The company will also publish the benchmark scores for the 27B variant, providing a clearer picture of its suitability for local deployment. Monitoring how the community adopts and modifies these models will be key to assessing their impact on the AI landscape in the coming months.
Building MCP Servers for AI Agents: Scalable Architecture Patterns, Security Design, and Production-Ready AI Infrastructure for Large Language Models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
When will Alibaba release the open weights for Qwen3.8-Max?
The open weights are scheduled for release next week, with the exact date to be announced by Alibaba.
How does Qwen3.8-Max compare to other models like GPT-5.6 or Fable 5?
Qwen3.8-Max ranks second to Fable 5 on several benchmarks, with performance close to GPT-5.6 at maximum effort, but trails notably in deep software-engineering tasks.
What are the licensing terms for the open weights?
The licensing details are still unpublished, so it is unclear whether the open weights will be freely available or subject to restrictions.
Will the smaller 27B model be as capable as the flagship?
The performance of the 27B variant remains unconfirmed, but historically, mid-sized models tend to be less capable in complex tasks compared to larger counterparts.
What does this development mean for the AI industry?
This marks a move toward greater transparency and openness in large language models, potentially influencing industry standards and encouraging broader access to powerful AI tools.
Source: ThorstenMeyerAI.com
