Kimi K2 Thinking API via TokenMix
Use Kimi K2 Thinking from Moonshot as a chat model through the TokenMix AI API relay and multi-model gateway.
Deep reasoning model with general agentic capabilities built on the K2 MoE architecture. Supports up to 300 steps of complex tool invocation for multi-step problem solving.
API access
- Base URL:
https://api.tokenmix.ai/v1 - Model ID:
kimi-k2-thinking - OpenAI SDK compatible. Change the base URL and use your TokenMix API key.
Pricing
Input $0.529412/M tokens, output $2.117647/M tokens
Capabilities
Function calling, JSON mode, Streaming, Reasoning
Model specs
- Context: 262K tokens
- Max output: 66K tokens
Availability
3/3 available API endpoints are healthy right now.
Recent performance
TTFT 374ms, latency 3712ms, throughput 67.7 tok/s.
Start using this model
Create an API key, top up from $1 when needed, and call this model through the TokenMix OpenAI-compatible endpoint.