No wonder Qwen and Gemma are so different
英文标题：No wonder Qwen and Gemma are so different
Pasted the same HTML/JS code (330 lines) into Qwen 35B A3B and Gemma 26B A4B. Qwen: tokenized the input to 1609 tokens Gemma: tokenized the input to 4258 tokens. Damn. I've never noticed this before and I haven't seen people mention it. That alone helps explain why Qwen is regarded as better at codi
打开原文相关 2 · 机会 0
中文
Pasted the same HTML/JS code (330 lines) into Qwen 35B A3B and Gemma 26B A4B. Qwen: tokenized the input to 1609 tokens Gemma: tokenized the input to 4258 tokens. Damn. I've never noticed this before and I haven't seen people mention it. That alone helps explain why Qwen is regarded as better at codi
中文全文翻译正在补齐，可先查看下方英文原文。
English / 原文
Pasted the same HTML/JS code (330 lines) into Qwen 35B A3B and Gemma 26B A4B. Qwen: tokenized the input to 1609 tokens Gemma: tokenized the input to 4258 tokens. Damn. I've never noticed this before and I haven't seen people mention it. That alone helps explain why Qwen is regarded as better at codi