Kimi K2 pricing for security review

Quick start

Call Kimi K2 with the OpenAI Python SDK pointed at api.moonshot.cn:

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.moonshot.cn/v1",
    api_key=os.environ["MOONSHOT_API_KEY"],
)
response = client.chat.completions.create(
    model="kimi-k2",
    max_tokens=2500,
    messages=[{"role": "user", "content": "Your prompt here"}]
)
# Cost: ~$0.0157 per call at standard rates
# Input:  $0.95/MTok
# Output: $4.00/MTok

How Kimi K2 compares on cost

Per-call cost for all models scoring 50+ on security review benchmarks. Hover for details.

Highest-quality affordable option: Claude Sonnet 5 at $0.0370/call (2.4x the cost, 96/100 quality). Kimi K2 is already the cheapest choice for security review.

What drives security review cost

Security review is the code task where quality is worth paying for: a missed vulnerability costs more than the entire year's inference bill. Input is large (the code plus the threat context) and output is moderate. This is the clearest case on this list for choosing a top-ranked model over the cheapest viable one, and for not batching away the latency you need in CI.

For Kimi K2 specifically, that means output tokens cost 4.2x more than input, at $4.00 vs $0.95 per MTok. At this use case’s 6,000/2,500 token split, output accounts for 64% of the $0.0157 per-call cost.

Kimi K2 quality benchmarks for security review

Benchmark data for security review is pending for Kimi K2. Test against your own evaluation suite before moving to production.

Benchmark data is not yet available for Kimi K2. We recommend testing against your own evaluation criteria.

About Kimi K2

Prices from platform.kimi.ai (K2.6, read 2026-10-11). Cache hit: $0.16/MTok input. Batch API at 60% of standard rate. Not yet benchmarked, so never recommended.

Other models from Moonshot (Kimi)

You can compare Kimi K2 head-to-head against any other model on our model comparison pages, or track how pricing has changed over time on the pricing history tracker.

Frequently asked questions