Moonshot AI·Kimi

Kimi K3

A 2.8T open MoE using KDA and Attention Residuals with native vision and a 1M context window.

CurrentReasoningCodingAgentMultimodalOpen weights

MODEL NUMBER025

Specifications

Only published specifications are shown.

Total parameters

2.8T

Architecture

MoE

Context

Input modalities

Text · Image

Open weights

Yes

Status

Current

A 2.8T open MoE using KDA and Attention Residuals with native vision and a 1M context window.

K3 pushed open-model scale toward 3T parameters while unifying agent, vision, and long-context capabilities.

Compared with Kimi K2 ↗

PreviousKimi K21T · 128K ctx

CurrentKimi K32.8T · 1M ctx

New capabilities Native vision Long-horizon agent Reasoning

Kimi Delta AttentionAttention Residuals

Native visionLong-horizon agentCodingReasoning

2026-07-16Announcement

Kimi K3 was announced at 2.8T parameters with native vision.

Select a card to view the related model.