Kimi K3, and what we can still learn from the pelican benchmark
Moonshot AI announces Kimi K3, a 2.8-trillion-parameter flagship with an open-weight release promised by July 27 — the first 'open 3T-class' model. Third-party evaluation (Artificial Analysis) puts it behind only Claude Fable 5 on long-horizon knowledge work, at a cost per task comparable to GPT-5.6 Sol.
Marked important, but published more than 30 days ago, so it moved down to this band.
On this page12 sections
The editors' notes below — why it matters, the context, the learning path — are written in Chinese.
Record
The signal
#Moonshot AI announced Kimi K3, described as their most capable model to date at 2.8 trillion parameters, available via their website and API with an open-weight release promised by July 27, 2026 — which would make it the first 'open 3T-class' model, taking the scale crown from DeepSeek's 1.6T v4 Pro. Self-reported benchmarks have K3 mostly beating Claude Opus 4.8 max and GPT-5.5 high while losing to Claude Fable 5 and GPT-5.6 Sol. Artificial Analysis reports a long-horizon knowledge-work Elo of 1547 (+732 over K2.6, behind only Claude Fable 5), cost per task of $0.94 (similar to GPT-5.6 Sol's $1.04, about half of Opus 4.8's $1.80, higher than open-weight peers), and 21% fewer output tokens than K2.6. The model also now leads Arena.ai's Frontend Code arena.
Why now
Why it matters
#开源权重前沿竞赛可能迎来新王座：若 7 月 27 日如约开放权重，K3 将是首个「3T 级」开放模型，且单任务成本已与 GPT-5.6 Sol 同档。闭源与开源的能力差距进一步收窄，「模型选型的默认答案」需要重新计算——尤其对可以自托管的团队。
Background
Technical context
#读这类发布要区分三层证据：厂商自报基准（普遍被针对性优化）、第三方系统评测（Artificial Analysis 的长时程 Elo 与单任务成本）、以及难以刷分的民间测试（Willison 的鹈鹕骑车基准）。K3 的输出 token 用量下降 21% 说明前沿竞争的重心正从纯能力转向单位算力与单位成本的智能密度——与 GPT-5.6 的主打方向相同。官方尚未详述架构细节（如是否 MoE、激活参数量）。
Audience
Who should care
#Next
Learning path
#- 先读「模型选型与约束匹配」，理解单任务成本比每 token 价格更接近真实决策变量
- 对照 Inkling（开源万亿 MoE）理解开放权重阵营的规模化路径
- 用一个自己领域的小评测集交叉验证自报基准与第三方数据
Network
Where it sits in the network
#How this signal connects to neighbouring technologies, skills and background knowledge. Follow any link to keep exploring.
Further thought
Follow-up questions
#- 开放权重是否在 7 月 27 日如期兑现？采用什么许可证？
- 架构细节（是否 MoE、激活参数量）公布后，自托管的真实算力门槛是多少？
- 开放后第三方复测能否复现自报与 Artificial Analysis 的结果？
Publication record
Featured in
#The briefings that carried this signal, newest first; the last row is the weekly review of its week. The tag is the section it ran in.
技术简报 - 2026-07-28
Immediate attentionWeekly review · Jul 13 – Jul 19
Worth tracking
Sources