Kimi K3.1 Pro, MiniMax M3, DeepSeek Flash and 4 more models: pricing and usage

Seven text models are now live on LaoZhang API: Kimi K3.1 Pro and its high-speed variant, MiniMax M3 and M3.1 Flash, two Zhipu GLM models, and deepseek-flash, DeepSeek’s new official model name. All seven use the OpenAI-compatible POST /v1/chat/completions endpoint and are billed per token, so existing code only needs a new model name.

Published: October 25, 2026
Last checked: October 25, 2026
Status: Live in the default group

At a glance

Item	Details
New model IDs	`kimi-k3.1-pro`, `kimi-k3.1-pro-highspeed`, `MiniMax-M3`, `MiniMax-M3.1-Flash`, `glm-5.2-fast-preview`, `glm-5.3-prime`, `deepseek-flash`
Endpoint	Base URL `https://api.laozhang.ai/v1`; Chat Completions `POST /v1/chat/completions`, streaming supported
Token group and billing	`default` group; set the billing mode to Usage first or Usage-based
Reasoning	All seven return their reasoning in `choices[0].message.reasoning_content`
Prices	See the table below; Models and pricing always shows the current rates

Pricing

USD per 1M tokens, from the public price list on October 25, 2026.

Model	Input	Output	Cache read
`kimi-k3.1-pro`	$2	$10	$0.2
`kimi-k3.1-pro-highspeed`	$4	$20	$0.2
`MiniMax-M3` (input ≤ 512K)	$0.3	$1.2	$0.06
`MiniMax-M3` (input > 512K)	$0.6	$2.4	$0.12
`MiniMax-M3.1-Flash`	$0.143	$0.571	$0.014
`glm-5.2-fast-preview`	$2.286	$8	—
`glm-5.3-prime`	$2.286	$8	—
`deepseek-flash`	$0.44	$1.32	$0.088

Kimi cache writes: kimi-k3.1-pro costs $2 for 5 minutes and $4 for 1 hour; kimi-k3.1-pro-highspeed costs $4 and $8.
When a MiniMax-M3 request has more than 512K input tokens, the whole request is billed at the higher tier.
deepseek-flash costs the same as deepseek-v4-flash.

Example

Create an API key under Token management in the default group, with the billing mode set to Usage first (recommended) or Usage-based. Then set it in your shell:

export LAOZHANG_API_KEY="your LaoZhang API key"

The examples use kimi-k3.1-pro; for the other six models, change only model. For Python, run python -m pip install openai first.

cURL
Python

curl https://api.laozhang.ai/v1/chat/completions \
  -H "Authorization: Bearer $LAOZHANG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3.1-pro",
    "messages": [
      {"role": "user", "content": "Explain a database index in three sentences."}
    ]
  }'

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["LAOZHANG_API_KEY"],
    base_url="https://api.laozhang.ai/v1",
    timeout=300.0,
)
response = client.chat.completions.create(
    model="kimi-k3.1-pro",
    messages=[{"role": "user", "content": "Explain a database index in three sentences."}],
)
message = response.choices[0].message
print(getattr(message, "reasoning_content", None))  # reasoning
print(message.content)                               # final answer

The final answer is in choices[0].message.content, and the reasoning is in choices[0].message.reasoning_content.
Token usage and charges for each request are in the usage log.

Who is affected

For: projects that want Kimi K3.1 Pro, MiniMax M3, the new GLM models, or DeepSeek’s new model name. Any API key in the default group with a usage-based billing mode can call them.
Not affected: no existing model changed. The earlier Kimi, MiniMax, GLM and DeepSeek models (for example kimi-k3 and deepseek-v4-flash) keep working, and existing integrations need no changes.

FAQ

What is the difference between `deepseek-flash` and `deepseek-v4-flash`?

They are the same model at the same price. DeepSeek now uses deepseek-flash as its model name. The old name deepseek-v4-flash still works and is served by DeepSeek-V4.1-Flash as well, so projects already using it need no changes.

Should I use the Kimi high-speed variant?

kimi-k3.1-pro-highspeed is Kimi’s high-speed version. Input and output cost twice as much as the standard model; cache reads cost the same. If latency matters, compare both with a few representative requests from your own workload before paying the higher rate.

Is the reasoning billed?

Yes. Reasoning tokens count as output tokens and are billed at the output rate. All seven models currently return reasoning_content; if you only need the answer, read content.

How are `MiniMax-M3` requests over 512K tokens billed?

When one request has more than 512K input tokens, its input, output and cache reads are all billed at the higher tier ($0.6 input, $2.4 output, $0.12 cache read), not just the part above 512K.

Where can I check current prices and actual charges?

Rates are on Models and pricing. The model, token usage and charge for each request are in the usage log.

Kimi API pricing: prices and cache billing for Kimi K3.1 Pro and the high-speed variant.
MiniMax pay-as-you-go pricing: the 512K pricing tiers for MiniMax-M3 and the M3.1-Flash price.
DeepSeek models and pricing: the deepseek-flash model name and the legacy deepseek-v4-flash name.
Models and pricing
Latest announcements

​At a glance

​Pricing

​Example

​Who is affected

​FAQ

​What is the difference between deepseek-flash and deepseek-v4-flash?

​Should I use the Kimi high-speed variant?

​Is the reasoning billed?

​How are MiniMax-M3 requests over 512K tokens billed?

​Where can I check current prices and actual charges?

​Sources and related pages