@kimmonismus Crazy how the compute bottleneck shifted so fast. If Kimi K3 needs cluster level infra just to run inference, open source scaling is hitting a wall.
@kimmonismus China keeps leading with big open models like Kimi K3. Good time for OpenAI to go open source too.
@kimmonismus The open source race is making the AI landscape much more competitive.
@kimmonismus Keeping everything proprietary just slows down the real progress for the community
Positive users praise Kimi K3 exceeding single NVIDIA B200 capacity and AMD MI455X memory fit for boosting open-source competition and performance, while negative users worry proprietary models slow progress and make home setups obsolete.
Based on 15 visible X reactions from 27 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@kimmonismus Crazy how the compute bottleneck shifted so fast. If Kimi K3 needs cluster level infra just to run inference, open source scaling is hitting a wall.
@kimmonismus China keeps leading with big open models like Kimi K3. Good time for OpenAI to go open source too.
@kimmonismus The open source race is making the AI landscape much more competitive.
@kimmonismus Keeping everything proprietary just slows down the real progress for the community
@arankomatsuzaki that is such a huge upgrade for performance honestly
@kimmonismus Kimi K3 es lo que debió ser chat gpt 5
The release of Kimi K3 would be a good time to release GPT-OSS-2, although I don't believe it will happen. The real Open AI currently comes from China, and that won't change for the time being. That should at least give us something to think about.
@arankomatsuzaki that is such a huge upgrade for performance honestly
@kimmonismus Kimi K3 es lo que debió ser chat gpt 5
The release of Kimi K3 would be a good time to release GPT-OSS-2, although I don't believe it will happen. The real Open AI currently comes from China, and that won't change for the time being. That should at least give us something to think about.
Good news / trend for AMD GPUs and GB / VR
The release of Kimi K3 would be a good time to release GPT-OSS-2, although I don't believe it will happen. The real Open AI currently comes from China, and that won't change for the time being. That should at least give us something to think about.
Kimi K3 2.8T is so large that it will not fit on a single NVIDIA DGX B200, even at FP4. A GB300 NVL72, B300, or MI355X system is required, as each GPU has 288 GB of memory. One optimization that could make Kimi K3 fit on B200 is to gang multiple nodes together and use a technique called WideEP. The issue is that B200 has only 400 Gbit/s of bandwidth between nodes, whereas NVL72 has 18× higher inter-node bandwidth.
Good news / trend for AMD GPUs and GB / VR
Positive users praise Kimi K3 exceeding single NVIDIA B200 capacity and AMD MI455X memory fit for boosting open-source competition and performance, while negative users worry proprietary models slow progress and make home setups obsolete.
Based on 15 visible X reactions from 27 accounts; directional sample.
Ask a question below.
Published answers will appear here.
Good news / trend for AMD GPUs and GB / VR
