kimik3.io/Try it free

How to try Kimi K3 for free

You can try Kimi K3 for free: chat on Moonshot’s official app, or call the API with the 10 free credits EvoLink adds at signup. Below: how the two routes compare, and every step to a first call.

You can try Kimi K3 today without paying. Route one: kimi.com — Moonshot's own web and mobile app, which officially carries K3 (per the quickstart, read 2026-07-18). Route two, for developers: the API via EvoLink, which gives 10 free credits at signup — enough to make your first real kimi-k3 calls with the verified code below.

Model availability per the official quickstart, read 2026-07-18. Paid rates, when you outgrow free: the official numbers with a date stamp.

Pick your route

What each free route gets you. For kimi.com's free-tier limits, check Moonshot directly — they change often; the API-route facts are ones we've verified.
Chat: kimi.comAPI: EvoLink free credits
Best forSeeing what K3 can do, hands onTesting K3 in your code and workload
What you getMoonshot's own app, web + mobile10 free credits at signup, OpenAI-compatible endpoint
SetupSign in and typeOne key + the quickstart
Vision / filesApp features, as Moonshot ships themAPI parameters — what we've verified
When it endsMoonshot's tier limitsCredits run out → official per-token rates, no markup

The developer route, start to finish

  1. Create an EvoLink account — 10 free credits land on signup.
  2. Create an API key in the dashboard and export it as EVOLINK_API_KEY.
  3. Run this — it's the exact code we verified against production the day K3 launched:
python · openai sdk
from openai import OpenAI

client = OpenAI(
    api_key="YOUR_EVOLINK_API_KEY",
    base_url="https://direct.evolink.ai/v1",   # <-- the one line you change
)

response = client.chat.completions.create(
    model="kimi-k3",
    messages=[{"role": "user", "content": "Explain prompt caching in two sentences."}],
)

print(response.choices[0].message.content)

Two settings protect your free credits: leave max_completion_tokens at its default (capping it can bill you for empty responses), and remember reasoning bills as output — 43–77% of output tokens in our runs. Auth, streaming, and a real response body: the API guide.

Make the free credits go further

  • Repeat your prefix. Caching is automatic and cuts repeated input from $3.00 to $0.30 per 1M tokens — the discount applies to trial credits the same as to paid usage.
  • Start with short prompts. At $3.00/1M input, prompt size is what eats a budget: a 497,718-token prompt costs $1.49 per callmeasured. Bring in long context once you need it.
  • Ask for short answers in the prompt, not via max_completion_tokens — the reliable way to keep output (and its reasoning overhead) small.
  • Know what "expensive" is before you commit: the calculator turns your real workload into a monthly number at official rates, in your browser.

Free-tier FAQ

Is Kimi K3 free to use?

To chat: Moonshot's own app at kimi.com carries K3 (official quickstart, read 2026-07-18); see Moonshot's own terms for tier limits. To build: the API is pay-per-token at $3.00/1M input and $15.00/1M output. You can go direct through Moonshot or through a third-party aggregator: EvoLink, for example, carries K3 at the same rates, one key also calls GPT, Claude, and Gemini, and signup includes 10 free credits — enough to verify K3 on your own workload before spending anything.

Is there a Kimi K3 API free trial?

EvoLink gives 10 free credits on signup. When they're used up, you pay the official per-token rates with no markup. Moonshot hasn't published a free API tier; if one appears, this page gets updated and the changelog records it.

Do I need a Chinese phone number to try Kimi K3?

Not via the EvoLink route — signup works from anywhere. One key also reaches GPT, Claude, Gemini and other mainstream models on the same OpenAI-compatible endpoint, so a comparison test costs nothing extra to set up.

What's the cheapest way to keep using Kimi K3 after the trial?

Structure your prompts for the automatic cache: stable prefix first, varying content last. Input drops 10× on hits — in our measured workloads that's the difference between $3.00 and $0.30 per million tokens on the biggest line item.

Claim the 10 free credits

EvoLink carries kimi-k3 on an OpenAI-compatible endpoint — one key reaches GPT, Claude, Gemini, and other mainstream models. 10 free credits at signup.

Get an EvoLink API key