Kimi K2 Thinking API: Amazon Bedrock (US)'s chat / LLM model, one key away
Kimi K2 Thinking via the ApexApi gateway
The reasoning variant of Kimi K2, which works through a problem before answering. Cheaper on output than K2.5 and aimed at analysis rather than chat. Call it through ApexApi's unified, OpenAI-compatible API with one `ak-` key, alongside every other model in the catalog. It supports a 256,000-token context window, tool / function calling, and streaming responses. Pricing is $0.72 input and $3.00 output per 1M tokens, billed pay-per-use from your credit balance with no subscription.
Kimi K2 Thinking is served through ApexApi's unified, OpenAI-compatible API. One key, one format, with smart routing and automatic failover. Provider: Amazon Bedrock (US).
Why Kimi K2 Thinking
256,000-token context
Reason over long documents, codebases, and multi-turn history in a single request.
Tool & function calling
Structured outputs and tool orchestration for agentic and automation workloads.
OpenAI-compatible
Drop-in with any OpenAI SDK. Change the base URL and key, keep your code.
Specifications
- Modality
- chat
- Context window
- 256K tokens
- Max output
- —
- Streaming
- Yes
- Tool / function calling
- Yes
- Vision input
- —
Kimi K2 Thinking: measured reliability
From real calls ApexApi made to Kimi K2 Thinking during health sweeps over the last 30 days. Our own measurements, not vendor claims.
- Success rate
- 100%
- Avg latency
- 9.4s
- Calls measured
- 1
- Last checked
- Aug 30
See how Kimi K2 Thinking compares against every model we serve on the model rankings.
Kimi K2 Thinking pricing
- A million tokens in and a million out costs $3.72 here, all-in.
- That makes it the 38th cheapest of the 81 chat models we serve, ranked on output price.
Getting started with Kimi K2 Thinking
OpenAI-compatible. Point your client at ApexApi, swap your key, and call it. No new SDK to learn.
from openai import OpenAI
client = OpenAI(
base_url="https://api.apexapi.dev/v1",
api_key="<YOUR_APEXAPI_KEY>",
)
response = client.chat.completions.create(
model="moonshotai/kimi-k2-thinking",
messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)Kimi K2 Thinking use cases
What teams build with Kimi K2 Thinking on ApexApi.
Agents & automation
Tool-using agents that orchestrate multi-step workflows and call APIs.
Coding assistants
Code generation, refactoring, reviews, and multi-file reasoning.
RAG & knowledge apps
Summarization and long-document Q&A grounded in your own data.
Kimi K2 Thinking vs other models
How it compares, and on ApexApi you can switch between any of them with a single string change, same key, same endpoint.
Kimi K2 Thinking vs Claude Opus 4.7
Compare Kimi K2 Thinking against this model on price, capabilities, and latency. A/B them with a single key on ApexApi.
Learn more about Claude Opus 4.7 API →Kimi K2 Thinking vs GPT-5.5
Compare Kimi K2 Thinking against this model on price, capabilities, and latency. A/B them with a single key on ApexApi.
Learn more about GPT-5.5 API →Kimi K2 Thinking vs Gemini 3.1 Pro Preview
Compare Kimi K2 Thinking against this model on price, capabilities, and latency. A/B them with a single key on ApexApi.
Learn more about Gemini 3.1 Pro Preview API →Kimi K2 Thinking API: FAQ
How do I call Kimi K2 Thinking through ApexApi?
Create an ApexApi key, point your client at https://api.apexapi.dev/v1, and POST to /v1/chat/completions with the model set to `moonshotai/kimi-k2-thinking`. The API is OpenAI-compatible, so an existing OpenAI SDK works once you change the base URL and key.
How much does Kimi K2 Thinking cost on ApexApi?
$0.72 per million input tokens and $3.00 per million output tokens, all-in with no separate platform fee. You buy credits and pay per call, with no subscription or minimum.
What is the context window for Kimi K2 Thinking?
256,000 tokens.
Does Kimi K2 Thinking support tool calling and image input?
Kimi K2 Thinking supports tool / function calling and streaming responses, and it does not support image (vision) input.
Ready to build with Kimi K2 Thinking?
One API key. Every major AI model. Pay only for what you use.
Get your API key, free to start