Microsoft announced on July 28 that users can deploy Moonshot AI's Kimi K3 model on Microsoft's Microsoft Foundry through Fireworks AI.
Note: Microsoft Foundry is Microsoft's enterprise-grade, one-stop AI application and agent development platform. It integrates model management, agent building, security governance, and observability into a single cloud-native architecture, aiming to help developers and enterprises build, run, and scale generative AI applications at scale.
Fireworks AI is a cloud services platform focused on high-speed, cost-effective generative AI inference and fine-tuning. It was founded in 2022 by former core members of Meta's PyTorch team (CEO Che Lin was formerly the global head of PyTorch) and is headquartered in California, USA.
Microsoft officially stated that users can deploy the Kimi K3 model on Microsoft Foundry through Fireworks AI. The model has 2.8 trillion parameters, making it one of the largest open models currently available.
Its model context window is 1 million tokens, meaning that a single request can process approximately 750,000 words (roughly equivalent to 10 complete novels or an entire codebase).
Model Name
Deployment Type
Input per 1 million Tokens
Output per 1 million Tokens
Cached Input per 1 million Tokens
FW Kimi K3
Data Zone
3.30 USD (approximately 22.4 yuan at the current exchange rate)
16.50 USD (approximately 111.8 yuan at the current exchange rate)
0.33 USD (approximately 2.2 yuan at the current exchange rate)