Running Kimi K3 on MI355X at Better Performance per Dollar Than B300
This article discusses the performance of the Kimi K3 AI model on AMD MI355X hardware, arguing that it offers superior performance-per-dollar compared to NVIDIA's B300 and B200 chips. It highlights the engineering challenges of serving massive 2.8T parameter models and suggests AMD is becoming a viable alternative for large-scale AI inference.
Why it matters
It highlights the shifting competitive landscape in AI hardware, where cost-efficiency and memory capacity are challenging NVIDIA's market dominance.
Running Kimi K3 at ~952 tok/s/node, AMD continues to prove its case as the winner in performance per dollar.
Over the past several months, we’ve seen an explosion in the capabilities of open source models. With DeepSeek V4-Pro and GLM5.2 reaching near-Opus levels of intelligence, open source has emerged as a real, cost-efficient alternative to the closed source models we’ve been married to.
But we have yet to see one like Kimi K3. Promising Fable/Sol levels of intelligence, Kimi K3 marks the start of a new era for open source.
Get smarter about the news
Sign up free for a feed built around what you actually care about, Dive Deeper research on any story, and the full text of every article.
Create free accountAlready have an account? Sign in