Alibaba's Qwen 3.8-max and Meta's Muse-spark-1.2 Enter Top LMSYS Arena Ranks
In the 32nd week of 2026, Alibaba's Qwen 3.8-max and Meta's Muse-spark-1.2 (xHigh) debuted on the LMSYS Arena overall leaderboard at 6th and 4th place respectively. Additionally, Moonshot's Kimi-k3-max climbed to 6th place on the coding leaderboard, showcasing strong performance from Chinese AI models. The entry of these new models into the top tier of the LMSYS Arena highlights the intensifying competition between Chinese tech giants and Western AI labs. It demonstrates that Chinese models are becoming highly competitive in both general capabilities and specialized tasks like coding. Qwen 3.8-max achieved an Elo rating of 1497, tying with Anthropic's Claude-opus-4-6, while Meta's Muse-spark-1.2 (xHigh) scored 1498. Anthropic's Claude-fable-5 continues to hold the top spot on both the overall and coding leaderboards.
## BACKGROUND
LMSYS Chatbot Arena is a widely recognized crowdsourced open platform for evaluating LLMs using blind A/B testing. It calculates Elo ratings based on human preferences from side-by-side model comparisons, serving as a key benchmark for real-world LLM performance.