NEWKimi K3 is live — 2.8T params · native vision · 1M context
One API, every leading AI model

NexusFlow

Between question and answer, there is always a path. NexusFlow turns that uncertainty into one deliberate API for text, vision, image and video intelligence.

45+model options
2.8TKimi K3 flagship
4K HDRSeedance 2.0 video
Kimi K3New
Moonshot AI1M context
In ¥20 · Out ¥100
Qwen3.7 MaxFlagship
Tongyi Qianwen1M context
In ¥12 · Out ¥36
Qwen3 MaxStable
Tongyi Qianwen262K context
In ¥2.5 · Out ¥10
Qwen LongLong
Tongyi Qianwen10M context
In ¥0.5 · Out ¥2
Qwen3.6 PlusPopular
Tongyi Qianwen1M context
In ¥2 · Out ¥12
Qwen3.5 PlusBalanced
Tongyi Qianwen1M context
In ¥0.8 · Out ¥4.8
Qwen3.5 FlashFast
Tongyi Qianwen1M context
In ¥0.2 · Out ¥2
Qwen3.5 Omni PlusOmni
Tongyi Qianwen262K omni
In ¥7 · Out ¥40
Qwen3.5 Omni FlashOmni
Tongyi Qianwen262K omni
In ¥2.2 · Out ¥13.3
Qwen VL FlashVision
Tongyi Qianwen262K vision
In ¥0.15 · Out ¥1.5
Qwen Coder FlashCode
Tongyi Qianwen1M code
In ¥1 · Out ¥4
DeepSeek V4 ProReasoning
DeepSeek1M context
In ¥12 · Out ¥24
DeepSeek V3.2General
DeepSeek131K context
In ¥2 · Out ¥3
GLM 5.2Flagship
Zhipu AI1M context
In ¥8 · Out ¥28
Text Embedding V4Vector
Tongyi Qianwen8K vectors
¥0.5 / 1M input
Qwen Image MaxImage
Tongyi QianwenImage
per image
PixVerse V4.5Video
PixVerseAsync video
from ¥0.15/s
HappyHorse 1.0Video
Tongyi QianwenAsync video
from ¥0.9/s
Flagship LLM最新上线by 月之暗面 Moonshot AI

Kimi K3

Kimi 迄今能力最强的旗舰模型:2.8 万亿参数,基于 KDA 混合线性注意力与注意力残差架构,原生视觉理解 + 深度思考,100 万 token 上下文。面向长程编程、知识工作与推理场景——OpenAI 与 Anthropic 协议均可直接调用 kimi/kimi-k3

全球首个开源的 3 万亿级别模型。输入 ¥20/M、输出 ¥100/M、缓存命中低至 ¥2/M,与 Qwen、GLM、DeepSeek 共用同一个 API Key 与计费体系。

了解 Kimi K3API 文档
2.8T
万亿级参数
1M
Token 上下文窗口
1M
最大输出长度
¥2/M
缓存命中输入价
Model access

Every request begins as a choice

Route requests across chat, reasoning, long-context, image and video models without multiplying accounts, keys and invoices. The catalog below is loaded from the live model API.

Kimi K3Moonshot AI1Minput ¥20 / output ¥100 per 1M
Qwen3.7 MaxTongyi Qianwen1Minput ¥12 / output ¥36 per 1M
GLM 5.2Zhipu AI1Minput ¥8 / output ¥28 per 1M
DeepSeek V4 ProDeepSeek1Minput ¥12 / output ¥24 per 1M
Seedance 2.0Volcengine ArkAsync videofrom ¥0.44 / second
Platform

Designed for teams, not demos

Unified API

Use one OpenAI-compatible endpoint for chat, embeddings, image, video, Anthropic Messages and Responses API calls.

Billing Control

Pre-call balance checks, precise micro-cost ledger entries, API key-level usage and account-level transaction history.

Operational Guardrails

Rate limits, upload authorization, production-safe payment handling, provider routing and health monitoring foundations.

Developer Console

Create keys, test prompts, inspect usage, monitor latency and manage tickets without switching provider dashboards.

Developer workflow

Give the question a path to follow

  1. Create an account and generate a one-time API key
  2. Point your SDK to https://nexusflow.hk/v1
  3. Choose a model per request or test in Playground
  4. Track cost, latency, errors and rate limits in the console

Not all answers are equal.Choose the route before the reply.

Validate in Playground, then ship through the same model names, keys and billing path in production.

免费开始