alt Hacker NewsThe best open models such as Kimi 2.5 are about as smart today as the big proprietary models were one year ago. That's not "nothing" and is plenty good enough for everyday work.
Replies
Which takes a $20k thunderbolt cluster of 2 512GB RAM Mac Studio Ultras to run at full quality…
The article mentions https://unsloth.ai/docs/basics/claude-codex
I'll add on https://unsloth.ai/docs/models/qwen3-coder-next
The full model is supposedly comparable to Sonnet 4.5 But, you can run the 4 bit quant on consumer hardware as long as your RAM + VRAM has room to hold 46GB. 8 bit needs 85.
LOCAL models. No one is running Kimi 2.5 on their Macbook or RTX 4090.
Kimi K2.5 is fourth place for intelligence right now. And it's not as good as the top frontier models at coding, but it's better than Claude 4.5 Sonnet. https://artificialanalysis.ai/models
Having used K2.5 I’d judge it to be a little better than that. Maybe as good as proprietary models from last June?
> The best open models such as Kimi 2.5 are about as smart today as the big proprietary models were one year ago
Kimi K2.5 is a trillion parameter model. You can't run it locally on anything other than extremely well equipped hardware. Even heavily quantized you'd still need 512GB of unified memory, and the quantization would impact the performance.
Also the proprietary models a year ago were not that good for anything beyond basic tasks.