Meta Muse Spark 1.1 Outperforms GLM-5.2 in Coding Benchmarks at a Slightly Lower Cost
Meta's updated Muse Spark 1.1 model scored 71.3 on coding benchmarks, surpassing Zhipu AI's GLM-5.2 (68.8) while costing $0.26 per task — cheaper than its Chinese competitor.
Introduction: A New Round in the AI War
Meta has released an update to its coding-focused AI model — Muse Spark 1.1 — which significantly improved performance on coding benchmarks and defeated Chinese Zhipu AI's flagship model GLM-5.2 both in quality and price. According to Artificial Analysis, Muse Spark 1.1 scored 51 on the Intelligence Index, placing it alongside GLM-5.2, GPT-5.4, and GPT-5.6 Luna. The model gained 8 points in just three months, primarily in coding and agent-based intelligent work.
On the Coding Index, Muse Spark 1.1 scored 71.3, surpassing GLM-5.2's 68.8. It narrowly trails GPT-5.6 Luna at 71.4. Leaders remain GPT-5.6 Sol (77.4), Terra (76.7), and Claude Fable 5 (76.5).
Price War and Improvements
Muse Spark 1.1 costs $0.26 per task — significantly less than GLM-5.2 ($0.37) and GPT-5.4 ($0.89). The hallucination rate dropped from 73% to 38%. The context window quadrupled to 1 million tokens — 4x the previous version and matching major competitors.
The US-China AI Race
Muse Spark 1.1's release comes when Meta's AI platform access is restricted in China, while Zhipu AI's models are free or cheap. Meta competes on price despite higher underlying costs. The gap between open and closed models shrinks each quarter.
Conclusion
Muse Spark 1.1 shows Meta's AI strategy is working. Developers and businesses now have more choice in the AI tools market, which benefits the entire industry. The battle for AI supremacy is far from over, but each new release makes powerful AI more accessible.