6 ms·
Benchmarks show that open models are equal to SOTA closed ones but own experience and real world use shows the opposite. And I really wish they were closer, I r
by rdos 11mo ago
Benchmarks show that open models are equal to SOTA closed ones but own experience and real world use shows the opposite. And I really wish they were closer, I run GPT-OSS 120b as a daily driver
- vinhnx 11mo agoIt could be that inference remote providers has issue, hence the model could not show potential or rate limited. I also think the Moonshot could take more time and continue with K2.1 or something like with DeepSeek. [0] https://x.com/Kimi_Moonshot/status/1986754111992451337 https://x.com/Kimi_Moonshot/status/1986754111992451337