[KongchangAI]
Model

Kimi

Kimi is a large language model developed by Chinese AI company Moonshot AI (月之暗面). The model features long-text understanding and generation capabilities, supporting tasks including multi-turn dialogue, document analysis, and coding assistance. An extended context window is among its primary characteristics. Kimi provides services to individual users and developers, and can be integrated into third-party applications and development tools via an API.

Core Facts

Timeline (last 90 days)

Oct 28

Domestic open-source models such as Huawei, Kimi, MiniMax, and DeepSeek made a concentrated appearance at this year's Wuzhen conference.

Unverified50%
Oct 28

In the blogger's test, DeepSeek V4.1 Flash's output for a 3D model of a foldable phone was better than the same prompt output from Kimi's new model, but the folding animation had a minor issue where it briefly moved out of the frame after clicking.

Unverified50%
Oct 27

Kimi's new membership system removed the weekly limit for coding, but all features of the entire membership (Agent, PPT, deep research, Kimi Code) share the same monthly quota pool; the official information only gives multiples: PLUS 2x, PRO 4x, MAX 14x, ULTRA 28x.

Unverified50%
Oct 27

The uploader calculated the yearly total by multiplying the annual-payment monthly-equivalent price by 12: for the entry tier, Zhipu, Kimi, and MiniMax are approximately 991, 948, and 490 respectively; for the highest tier, approximately 9055, 6708, and 4690 respectively.

Unverified50%
Oct 27

For existing users of Kimi's old packages (such as Andante and Moderato), the official commitment is that the name, benefits, and automatic renewal remain unchanged, and the weekly limit is retained.

Unverified50%
Oct 27

Kimi Code is sold bundled with membership; the first tier, GOLD, does not include Kimi Code, so those who want to write code must start from PLUS; the annual PLUS works out to a lower average monthly price than monthly payment; Kimi Code also has a higher-priced Ultra tier.

Unverified50%
Oct 27

For monthly entry-tier pricing, the order is Zhipu highest, Kimi in the middle, and MiniMax lowest; the middle tier and the highest tier follow the same order.

Unverified50%
Oct 25

The domestic version of Coze has a Chinese interface and can use models such as Doubao, DeepSeek, Kimi and Jieyue Xingchen.

Unverified50%
Oct 25

Nathan Lambert points out that China has a large number of API proxies and model proxies with enormous token usage, and there are even cases such as Kimi offering Claude services to users within its product, which makes him keep error bars on the conclusion that "distillation is not a major factor."

Unverified50%
Oct 25

Coopanion supports 13 model providers including DeepSeek, Qwen, Kimi, and Claude, and users can switch models by providing an API Key.

Unverified50%

40 more timeline events

All Facts (20)

Verified

Kimi Code is a programming tool from Moonshot AI.

80%
Verified

Early versions of the Kimi series of models supported a context input of 2 million Chinese characters.

80%
Verified

Kimi was developed by Moonshot AI and is known for its ultra-long context processing capabilities.

75%
Verified

Moonshot AI's Kimi is known for its ultra-long context window, supporting up to 2 million tokens.

70%
Verified

DeepSeek, Qwen, and GLM all belong to the camp of open-source models led by Chinese teams.

65%
Verified

Moonshot AI's Kimi series models are known for their long-context capability, which supports extremely long text input.

65%
Verified

The K3 large model released by Kimi has a parameter scale of nearly 3 trillion (2.8 trillion).

65%
Unverified

Kimi 国内可直接使用,无需翻墙

95%
Unverified

TRAE内置了对DeepSeek、Kimi等国产大模型的原生支持

90%
Unverified

Kimi智能助手最初以支持20万字超长文本输入为卖点切入市场

85%
Unverified

DeepSeek和Kimi等国内知名AI开源项目托管在GitHub上

85%
Unverified

此前的AI做PPT三件套需要手动复制粘贴到Gamma、Kimi等工具完成最终视觉呈现

85%
Unverified

遇到Bug时,用户只需截图发给AI并用日常语言描述问题和期望效果,无需指定CSS属性或代码行号

85%
Unverified

Kimi模型系列最受称道的特性之一是其超长上下文处理能力

80%
Unverified

MUON优化器最早由月之暗面(Kimi)团队在其k1.5技术报告中公开提出

75%
Unverified

Kimi曾被指控使用Claude的输出进行训练,因为它偶尔会说出'嗨,我是Claude'这样的话

75%
Unverified

Kimi系列模型底层架构采用了混合专家模型(MoE)设计

70%
Unverified

WorkBuddy supports calling multiple large models, including Kimi and DeepSeek.

60%
Unverified

WorkBuddy supports custom model imports, allowing users to connect their preferred open-source models.

60%
Unverified

The MUON optimizer was first validated in the Kimi model.

60%

Source Articles