Turiloop
中文API Key
All posts

How to access Kimi K2.6 and GLM-5.1 from outside China

· Kimi· GLM· Guide

DeepSeek gets the headlines, but two other models from the April 2026 Chinese wave are worth your attention: Kimi K2.6 from Moonshot and GLM-5.1 from Z.ai. This guide covers what each is good at and how to call them from outside China.

What each model is for

Kimi K2.6 (Moonshot AI) is built for long-context and agentic work. It scores 66.7% on Terminal-Bench 2.0 and about 58.6% on SWE-Bench Pro, which puts it right behind the closed frontier on real agent tasks. It's also known for aggressive prompt-cache pricing (around $0.16/M on cache reads) that makes long, stable system prompts cheap to reuse.

GLM-5.1 (Z.ai) is a strong coding and tool-use model, shipped in the same April 2026 wave as DeepSeek V4 and Kimi K2.6. It's a well-balanced default for code generation and agent loops at a low price.

The overseas problem

Calling Moonshot or Z.ai directly from abroad means a mainland phone number for sign-up and Alipay or WeChat for billing, the same wall you hit with any Chinese provider. The fix is a relay that holds the upstream access and exposes an OpenAI-compatible endpoint you pay for with a card.

Calling them: OpenAI-compatible, one base URL

With Turiloop, point any OpenAI SDK at the gateway and pick the model. Kimi K2.6:

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["TURILOOP_API_KEY"],
    base_url="https://api.turiloop.com/v1",
)

resp = client.chat.completions.create(
    model="kimi-k2.6",
    messages=[{"role": "user", "content": "Summarize this 200-page contract."}],
)
print(resp.choices[0].message.content)

Switching to GLM-5.1 is a one-line change, same key and same code:

resp = client.chat.completions.create(
    model="glm-5.1",
    messages=[{"role": "user", "content": "Refactor this module and add tests."}],
)

Or with curl:

curl https://api.turiloop.com/v1/chat/completions \
  -H "Authorization: Bearer $TURILOOP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "glm-5.1", "messages": [{"role": "user", "content": "Hello"}]}'

What you'll pay

On Turiloop, metered per million tokens (input / output):

  • Kimi K2.6: cheaper than any closed frontier model, and discounted right now. Best when you need long context and agentic reliability.
  • GLM-5.1: a low-cost, balanced default for coding.

Both take the same international card payment, pay-as-you-go, no Chinese phone number.

Which to pick

  • Long documents, multi-step agents, heavy prompt reuse → Kimi K2.6.
  • Everyday code generation and tool calls on a budget → GLM-5.1.
  • Need both? Keep one key and switch the model field per request.

Create a key on Turiloop and call Kimi K2.6 and GLM-5.1 in minutes.