Qwen 3.8 Max Tops Claude Opus in Benchmarks

// 16h agoMODEL RELEASE

Qwen 3.8 Max Tops Claude Opus in Benchmarks

An early evaluation of Alibaba's Qwen 3.8 Max large language model reveals that it outperforms all Claude Opus models in raw intelligence benchmarks. While the evaluator highlights its impressive capabilities, they also note that the model currently faces tool-calling issues when evaluated with certain testing harnesses.

// ANALYSIS

Massive open-weights models are rapidly catching up to proprietary LLM leaders, but developer-side integration reliability remains a key differentiator.

* Qwen 3.8 Max demonstrates frontier-class reasoning and intelligence, posing a direct threat to leading proprietary models in pure benchmark tests.

* Consistent tool-calling execution is still a bottleneck across different harnesses, showing that raw intelligence does not immediately guarantee seamless agentic integration.

* The model's preview signals Alibaba's commitment to pushing the boundaries of open-weights AI at the multi-trillion parameter scale.

// TAGS

qwen-3.8-maxllmbenchmarksalibaba-cloudartificial-intelligence

DISCOVERED

16h ago

2026-07-19

PUBLISHED

16h ago

2026-07-19

RELEVANCE

8/ 10

AUTHOR

aicodeking

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

OPEN SOURCE59m ago

Bojie Li open-sources AI agent textbook

Authored by Bojie Li, former Chief Scientist at Pine AI, "ai-agent-book" is an open-source Chinese textbook covering the core design principles and engineering patterns of AI agents. The repository hosts the full text, a compiled PDF version, and chapter-by-chapter code implementations.

OPEN SOURCE4h ago

Open Science v0.4.1 transitions to infrastructure

AIPOCH has released Open Science v0.4.1, transitioning the project from a desktop application into a model-agnostic, extensible AI research infrastructure. The new version enables reproducible, inspectable workflows across varied environments to facilitate distributed scientific discovery.

NEWS4h ago

Anthropic's Claude Fable AI model has generated a verified counterexample disproving the 87-year-old Jacobian Conjecture.

On July 19, 2026, an Anthropic researcher using the newly released Claude Fable model announced a concrete counterexample to the Jacobian Conjecture, a famous open problem in algebraic geometry since 1939. The model constructed a polynomial map from three-dimensional space to itself that has a constant Jacobian determinant of -2 but is not injective, mapping three distinct points—(0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2)—to the same output, (-1/4, 0, 0). Because the map fails to be one-to-one, it cannot have a polynomial inverse, disproving the conjecture for all dimensions n >= 3.