Qwen3.6-35B-A3B on my laptop drew me a better pelican than Claude Opus 4.7

simonwillison.net

April 16, 2026

2 min read

🔥🔥🔥🔥🔥

67/100

Summary

Qwen3.6-35B-A3B generated a superior image of a pelican compared to Claude Opus 4.7. The Qwen model was run on a MacBook Pro M5 using LM Studio.

Key Takeaways

Qwen3.6-35B-A3B produced a better illustration of a pelican riding a bicycle than Claude Opus 4.7.
The pelican benchmark, originally intended as a joke, has shown a correlation between the quality of generated images and the overall usefulness of AI models.
Qwen3.6-35B-A3B is currently more effective for generating specific SVG illustrations compared to Claude Opus 4.7.
The performance of Qwen3.6-35B-A3B does not necessarily indicate it is more powerful or useful than Anthropic's Claude Opus 4.7.

Coverage & discussions

Hacker News · simonwillison.net

Historical count: 5 · 5 saved in sample · Historical sample

Analysis ready

Community Sentiment

Mixed

Positives

Qwen 3.6-35B-A3B demonstrated notable creativity in generating unique and imaginative outputs, suggesting potential for artistic applications in AI.
The slight improvement in coding tasks performance indicates ongoing advancements in model capabilities, even if marginal.

Concerns

Despite its creative outputs, Qwen 3.6-35B-A3B struggles with physical plausibility, raising concerns about its practical applications in real-world scenarios.
The performance gap between Qwen and more established models like Opus 4.6 highlights significant limitations in Qwen's capabilities, particularly in benchmark tasks.