5 ms·
Qwen 2.5 32B which is an older model at this point clearly outperforms it: https://llm-stats.com/models/compare/gpt-3.5-turbo-0125-vs-qwen-2.5-32b-instruct htt
by electroglyph 1y ago
Qwen 2.5 32B which is an older model at this point clearly outperforms it:
https://llm-stats.com/models/compare/gpt-3.5-turbo-0125-vs-qwen-2.5-32b-instruct https://llm-stats.com/models/compare/gpt-3.5-turbo-0125-vs-q...
- ls612 1y agoEven when quantized down to 4 bits to fit on a 4090?
- FuckButtons 1y agoNot in my experience, running qwen3:32b is good, but it’s not as coherent or useful as 3.5 at a 4bit quant. But the gap is a lot narrower than llama 70b.