@Dorialexander At 15$/m it's a regression. It's an inefficient model which forces them to price it high. Moonshot are pretenders after all. All hope for AI revolution is on the whale, Zai and Qwen open source releases (and PrismML too) now.
@yzhang_cs @nathancgy4 @Xinyu2ML @Yulun_Du huge congrats! absolute banger model 🙌🏻
@yzhang_cs @nathancgy4 @Xinyu2ML @Yulun_Du Congrats! Thanks for the great contribution.
Many users congratulated the K3 open-source MoE model researchers because they saw the release after a year of work as an impressive technical achievement and valuable open-source contribution.
Based on 29 visible X reactions from 36 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@Dorialexander At 15$/m it's a regression. It's an inefficient model which forces them to price it high. Moonshot are pretenders after all. All hope for AI revolution is on the whale, Zai and Qwen open source releases (and PrismML too) now.
@yzhang_cs @nathancgy4 @Xinyu2ML @Yulun_Du huge congrats! absolute banger model 🙌🏻
@yzhang_cs @nathancgy4 @Xinyu2ML @Yulun_Du Congrats! Thanks for the great contribution.
@yzhang_cs @nathancgy4 @Xinyu2ML @Yulun_Du congratulations, this is amazing
> It is a monster model, yet clean in architecture; elegant in optimization, beautiful in MoE designs. I've been saying they do think of their project in aesthetic terms (also this looks like a V4 subtweet lmao)
The large-scale model was developed over a year-long project.
@yzhang_cs @nathancgy4 @Xinyu2ML @Yulun_Du congratulations, this is amazing
> It is a monster model, yet clean in architecture; elegant in optimization, beautiful in MoE designs. I've been saying they do think of their project in aesthetic terms (also this looks like a V4 subtweet lmao)
It does feel like a clicking moment for open source: K3 filling up the next scaling stage with something impressive but not fully finished, just as the post-training infra is stepping up.
It does feel like a clicking moment for open source: K3 filling up the next scaling stage with something impressive but not fully finished, just as the post-training infra is stepping up.
> It is a monster model, yet clean in architecture; elegant in optimization, beautiful in MoE designs. I've been saying they do think of their project in aesthetic terms (also this looks like a V4 subtweet lmao)
In the early morning, @nathancgy4, @Xinyu2ML, @Yulun_Du and I were preparing some showcases for the blog while watching the World Cup. The moment Argentina beat England, I felt something. I looked up, and saw the most unforgettable Beijing sunrise. So I took the photo. I knew this was no ordinary day, and now it comes and says, "Hello, world!" Happy hatch day, K3! Please enjoy it, I have felt 「眩晕瘫坐」 for weeks. It is a monster model, yet clean in architecture; elegant in optimization, beautiful in MoE desigins. It has crossed a certain singularity, it is a year of work, in one release. 🚀🚀🚀 This is a Sputnik moment for OSS.
In the early morning, @nathancgy4, @Xinyu2ML, @Yulun_Du and I were preparing some showcases for the blog while watching the World Cup. The moment Argentina beat England, I felt something. I looked up, and saw the most unforgettable Beijing sunrise. So I took the photo. I knew this was no ordinary day, and now it comes and says, "Hello, world!" Happy hatch day, K3! Please enjoy it, I have felt 「眩晕瘫坐」 for weeks. It is a monster model, yet clean in architecture; elegant in optimization, beautiful in MoE desigins. It has crossed a certain singularity, it is a year of work, in one release. 🚀🚀🚀 This is a Sputnik moment for OSS.
Introducing Kimi K3: Open Frontier Intelligence 🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal 🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts 🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost 🔹 Built for long-horizon agentic coding and self-evolving workflows Kimi K3 is now live on on http://Kimi.com, Kimi Work, Kimi Code, and the Kimi API. Open Weights by July 27, 2026. 🔗 API: http://platform.kimi.ai 🔗 Tech blog: http://kimi.com/blog/kimi-k3
In the early morning, @nathancgy4, @Xinyu2ML, @Yulun_Du and I were preparing some showcases for the blog while watching the World Cup. The moment Argentina beat England, I felt something. I looked up, and saw the most unforgettable Beijing sunrise. So I took the photo. I knew this was no ordinary day, and now it comes and says, "Hello, world!" Happy hatch day, K3! Please enjoy it, I have felt 「眩晕瘫坐」 for weeks. It is a monster model, yet clean in architecture; elegant in optimization, beautiful in MoE desigins. It has crossed a certain singularity, it is a year of work, in one release. 🚀🚀🚀 This is a Sputnik moment for OSS.
Introducing Kimi K3: Open Frontier Intelligence 🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal 🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts 🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost 🔹 Built for long-horizon agentic coding and self-evolving workflows Kimi K3 is now live on on http://Kimi.com, Kimi Work, Kimi Code, and the Kimi API. Open Weights by July 27, 2026. 🔗 API: http://platform.kimi.ai 🔗 Tech blog: http://kimi.com/blog/kimi-k3
In the early morning, @nathancgy4, @Xinyu2ML and I were preparing some showcases for the blog while watching the World Cup. The moment Argentina beat England, I felt something. I looked up, and saw the most unforgettable Beijing sunrise of my life. So I took the photo. I knew this was no ordinary day, and now it comes and says, "Hello, world!" Happy hatch day, K3! Please enjoy it, I have felt 「眩晕瘫坐」 for weeks. It is a monster model, yet clean in architecture; elegant in optimization, beautiful in MoE desigins. It has crossed a certain singularity, it is a year of work, in one release. 🚀🚀🚀 This is a Sputnik moment for OSS.
Introducing Kimi K3: Open Frontier Intelligence 🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal 🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts 🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost 🔹 Built for long-horizon agentic coding and self-evolving workflows Kimi K3 is now live on on http://Kimi.com, Kimi Work, Kimi Code, and the Kimi API. Open Weights by July 27, 2026. 🔗 API: http://platform.kimi.ai 🔗 Tech blog: http://kimi.com/blog/kimi-k3
Many users congratulated the K3 open-source MoE model researchers because they saw the release after a year of work as an impressive technical achievement and valuable open-source contribution.
Based on 29 visible X reactions from 36 accounts; directional sample.
Ask a question below.
Published answers will appear here.
It does feel like a clicking moment for open source: K3 filling up the next scaling stage with something impressive but not fully finished, just as the post-training infra is stepping up.
