Qwen 4-Flash Surpasses Opus 5.5 in Game Development Benchmark

Alibaba's Qwen4 series has shown benchmark results surpassing Opus5.5 in a game development creativity evaluation, according to news.aibase.com.

Benchmark Results

The GameFeel Bench, a creativity assessment focused on game development, showed Qwen4-Flash scoring 49.9 points, surpassing Opus5.5xHigh, which scored 47.2. The smaller, locally deployable Qwen4-27B scored 44.6, performing almost as well as Sonnet5.5, which scored 45.

Cost Comparison

According to leaked information cited by news.aibase.com, Qwen4-Flash costs about $0.4 to complete an equivalent task, far lower than the $2.6 required by Opus5.5. Qwen4-27B's execution cost is reduced to $0.16.

Qwen 4 Preview Capabilities

Geeky Gadgets reported on Qwen 4 preview capabilities, noting the model generated a Minecraft clone with features including shaders, a day-night cycle, an inventory system, water dynamics, underwater visuals, water breathing mechanics, a health bar, multiple terrains, islands, mobs, tree types, a cave system, and diamonds. The clone had minor flaws, including an inability to exit water and a distracting shader hue.

In another test, Qwen 4 created a detailed Roman coliseum with spectators, intricate architecture, atmospheric lighting, and flickering effects in "high thinking" mode, consuming about 69,000 tokens.

The Qwen 4 Curia checkpoint generated a 2D RPG village in a single HTML file, with superior composition compared to its GPT 611 Terra counterpart at maximum reasoning effort.

Model Lineup and Release

The Qwen 4 lineup includes Qwen 4 Max, Qwen 4 Flash, Qwen 4 Plus, and Qwen 4 27B, with traces of a specialized Qwen Coder model. Slides from the Esper conference confirmed that both the Qwen 4 and Qwen 5 series will remain open-weight.

Why it matters

The results indicate a Chinese large model has surpassed an overseas rival of this caliber in a related benchmark for the first time, while offering significantly lower inference costs.

---

Sources

• geeky-gadgets.com: Qwen 4 Preview First Look and Open Weight Capabilities - Geeky Gadgets
• news.aibase.com: Chinese AI Reaches a Highlight Moment: Qwen4 Benchmark Scores Surpass Opus 5.5 for the First Time, with Exceptional Cost-Effectiveness as Its Biggest Trump Card

This article was drafted with AI assistance and checked against the sources above. Company claims are reported as claims. Cover image is AI-generated.