tencent/kimi-k3

Common Name: Kimi K3

TencentCloud
SupportedTool InvocationSupportedReasoning
CompareTry in Chat

Kimi's flagship multimodal reasoning model for long-context knowledge work and agentic workflows.

Specifications

Context
1000K
Maximum Output
1000K
Inputtext, image, video
Outputtext

Performance (7-day Average)

Collecting…
Collecting…
Collecting…

Pricing

Input¥20.00/MTokens
Output¥100.00/MTokens
Cached Input¥2.00/MTokens

Performance Metrics (24h)

Similar Models

¥6.40/¥22.40/M-20%
ctx1.0Mmax128Kavailtps
InOutCap

GLM-5.2 is Zhipu AI's newest flagship model, served via Tencent Cloud, with 1M lossless context window. Released under the MIT open-source license, it delivers state-of-the-art performance on long-context engineering benchmarks (FrontierSWE 0.744, SWE-Bench-Pro 0.621).

¥6.00/¥18.00/M
ctx1.0Mmax64Kavailtps
InOutCap

Tencent's latest flagship reasoning model (770B/49B MoE) for coding agents and multi-step agentic workflows, with 1M context.

¥7.20/¥25.20/M-10%
ctx1.0Mmax128Kavailtps
InOutCap

Zhipu AI's latest flagship model, served via Tencent Cloud, with 1M context window and configurable thinking depth.

¥0.80/¥2.80/M
ctx1.0Mmax128Kavailtps
InOutCap

Zhipu AI's first natively multimodal GLM-5 series model (320B/18B MoE), served via Tencent Cloud TokenHub, with 1M context and visual coding capabilities.