A surprising contender has emerged in the AI coding assistant race. Recent benchmarks suggest Alibaba’s Qwen3-235B-A22B large language model is giving top-tier models like Anthropic’s Claude 3.7 (Opus) a serious run for their money, specifically within the Aider coding benchmark environment.
What’s turning heads is that Qwen3 achieved these results apparently without leveraging dedicated “thinking” processes that other models often employ.
The findings surfaced in a discussion surrounding a pull request on the GitHub repository for Aider, a popular AI pair programming assistant known for its helpful leaderboard comparing different models’ coding abilities. Initial tests submitted showed the Qwen3-235B model performing exceptionally well.
Interestingly, the Qwen team chimed in on the Aider discussion, suggesting slightly different settings for running their model in “non-thinking” mode.
They recommended parameters like a temperature of 0.7 and TopP of 0.8. When the benchmark was re-run with these updated configurations, Qwen3’s performance improved even further, jumping from an already impressive 61.8% to 65.3% on the Aider benchmark (whole format). This score puts it squarely in competition with, and in this specific test, seemingly ahead of Claude 3.7 Opus results which relied on thinking tokens.
This benchmark result is significant because it highlights the rapid advancements happening outside the most commonly discussed AI labs. While benchmarks are just one measure of performance, Qwen3’s strong showing on Aider, particularly with optimized settings, suggests it’s a powerful new option for developers looking for capable AI coding assistance. It underscores the value of community benchmarking efforts like Aider’s in revealing the real-world capabilities of emerging models.
- Forget Towers: Verizon and AST SpaceMobile Are Launching Cellular Service From Space
- This $1,600 Graphics Card Can Now Run $30,000 AI Models, Thanks to Huawei
- The Global AI Safety Train Leaves the Station: Is the U.S. Already Too Late?
- The AI Breakthrough That Solves Sparse Data: Meet the Interpolating Neural Network
- The AI Advantage: Why Defenders Must Adopt Claude to Secure Digital Infrastructure
