99% offKimi K2 Thinking API
Moonshot AI·Text· 100.0% success (7d)
Kimi K2 Thinking
Pricing
Cheapest live price, per 1 million tokens.
Input / 1M tokens
$0.600 OpenRouter
Output / 1M tokens
$2.50 OpenRouter
Cache read / 1M
Repeated prompt prefixes
Cache read is the discounted rate for repeated prompt prefixes (system prompts, agent context).
Capabilities · 6
Specifications
- Context window
- 256K
- Max output
- 66K
- Type
- Text
- Creator
- Moonshot AI
- Live sellers
- 188
- Requests / 24h
- 1,819
Price history · 60 days
Estimated monthly cost
30 days at the cheapest live price.
Benchmarks & reliability
LiveBench score + measured network performance. Leaderboard →
Quick start
OpenAI-compatible — change the base URL and key. Get a key from the dashboard.
from openai import OpenAI
client = OpenAI(
base_url="https://api.qubax.ai/v1",
api_key="YOUR_QUBAX_API_KEY",
)
response = client.chat.completions.create(
model="kimi-k2-thinking",
messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)About Kimi K2 Thinking
Kimi K2 Thinking is a language model developed by Moonshot AI. It's particularly well-suited for tackling complex, multi-step reasoning problems, orchestrating tool-using agents that call APIs and external services, producing structured outputs for reliable data extraction and handling large contexts of up to 256K tokens. Its 256K-token context window means you can pass in entire documents, codebases, or conversation histories without losing context. It can write up to 66K tokens in a single response.
As of Nov 4, 2026, Kimi K2 Thinking costs $0.0089 per 1M input tokens and $0.037 per 1M output tokens on Qubax, versus $0.600 / $2.50 on OpenRouter (99% less on input). A typical 1,000 chat requests (2K tokens in, 500 out) come to about $0.04. Measured availability over the last 7 days: 100.0%. 188 independent compute providers currently serve it, so requests fail over automatically. It handled 1,819 requests on Qubax in the last 24 hours. Pay with USDC, USDT, BTC, ETH or 350+ other coins — no credit card required.
Similar-priced alternatives
Frequently asked questions
How much does the Kimi K2 Thinking API cost?
On Qubax, Kimi K2 Thinking costs $0.0089 per 1M input tokens and $0.037 per 1M output tokens. This is 99% cheaper than OpenRouter's price of $0.600 per 1M input tokens.
Is the Kimi K2 Thinking API compatible with OpenAI?
Yes. Qubax provides an OpenAI-compatible API. You can use Kimi K2 Thinking as a drop-in replacement by changing your base URL to https://api.qubax.ai/v1 and using your Qubax API key. It works with Cline, Cursor, OpenCode, Hermes Agent, and any tool that supports OpenAI APIs.
Why is Kimi K2 Thinking cheaper on Qubax than OpenRouter?
Qubax sources AI inference from an open compute marketplace on the Base blockchain where GPU providers compete on price. We pass the savings to you — 99% cheaper than OpenRouter's price for Kimi K2 Thinking.
How do I pay for Kimi K2 Thinking API usage?
Qubax uses prepaid crypto credits. Top up with USDC, USDT, BTC, ETH, SOL or 350+ other coins and your balance is deducted as you use the API. No credit card or KYC required.
What is the context window of Kimi K2 Thinking?
Kimi K2 Thinking supports a context window of 256K tokens, allowing you to process large documents, codebases, or long conversations in a single request.
What can I build with Kimi K2 Thinking?
You can use Kimi K2 Thinking to tackle complex multi-step reasoning tasks, build autonomous agents that call external tools and APIs, generate structured JSON output for data pipelines and process long documents and entire codebases. Because it's served through Qubax's OpenAI-compatible API, integration into existing apps, agents, or workflows takes just a few lines of code.
How fast and reliable is Kimi K2 Thinking on Qubax?
Measured time to first token (p95) on Qubax is 3.1s. Its 7-day success rate is 100.0%. Responses stream token by token. 188 sellers serve it, so if one is slow or down the request moves to the next.
How good is Kimi K2 Thinking on benchmarks?
Kimi K2 Thinking scores 70.5 on LiveBench, ranking #12 of 16 models we track. At Qubax prices that is 4422.6 points per $1 per 1M tokens.
Kimi K2 Thinking vs GLM 4.5: which should I choose?
Kimi K2 Thinking by Moonshot AI and GLM 4.5 by Zhipu AI are both strong models in a similar price tier. Kimi K2 Thinking excels at complex reasoning. GLM 4.5 comes from a different lab and may have different strengths depending on your workload. On Qubax, Kimi K2 Thinking starts at $0.0089/1M input tokens while GLM 4.5 comes in at $0.0088/1M, so GLM 4.5 may be worth considering if you need to minimize per-token cost. Both are available on Qubax with crypto payments and OpenAI-compatible APIs, so you can try each and switch freely.
Can I use Kimi K2 Thinking for commercial projects?
Yes. Kimi K2 Thinking on Qubax can be used for commercial purposes. You own the outputs you generate. Qubax provides the API infrastructure — you bring your use case.
Why use Kimi K2 Thinking on Qubax?
99% cheaper
Providers compete on price — you get the cheapest healthy seller.
Pay with crypto
350+ coins (BTC, ETH, SOL, USDT, USDC). No card, no KYC.
OpenAI & Anthropic compatible
Drop-in base URL. Works with Cline, Cursor, Claude Code, OpenAI and Anthropic SDKs.