Pricing

Model multipliers, cache hits, and first API-key purchase

Model Pricing

Billing is based on 100M tokens: 1x costs CNY 40, or CNY 0.40 per 1M tokens. Purchase and top-up packages share the same six self-service tiers; contact support for larger volume.

Get API accessTop up existing keyRead integration docs
Existing usersAlready have an API key? Top it up directlyRecharge, order lookup, and delivery checks all happen here. New packages go as low as CNY 26.67 / 100M tokens.Go to top-up
Billing

How Billing Works

We bill in 100M-token units. Multipliers simply map model cost differences, and one key works across all supported models.

Core conversionStart with the baseline price
100M tokensCNY 40

Baseline at 1x

1M tokensCNY 0.40

Converted at 1x

USD-equivalentCNY 0.20

NewAPI $2 / 1M maps to our 1x standard

1M tokens cost CNY 0.40 at the baseline. A NewAPI marketplace price of $2 / 1M maps to our 1x standard.

One API key works across all models. Multipliers keep model costs easy to compare, so you can choose by use case without changing keys.

BaselineMultiplier Quick Reference
Multiplier100M tokens1M tokens
0.1xCNY 4CNY 0.04
0.5xCNY 20CNY 0.20
1xCNY 40CNY 0.40
2xCNY 80CNY 0.80
2.5xCNY 100CNY 1.00
3xCNY 120CNY 1.20
4xCNY 160CNY 1.60
5xCNY 200CNY 2.00
6xCNY 240CNY 2.40
7.5xCNY 300CNY 3.00
8xCNY 320CNY 3.20
10xCNY 400CNY 4.00
12xCNY 480CNY 4.80
15xCNY 600CNY 6.00
20xCNY 800CNY 8.00
100xCNY 4000CNY 40.00
Access URLPrefer the /v1 Base URL
Recommended Base URLhttps://gpt-agent.cc/v1

Use this first for most clients. If your client rejects it, switch to the fallback.

Fallback Base URLhttps://gpt-agent.cc

Different API routers validate Base URLs differently, so test the address when you switch models or clients.

Unified quotaOne key for all supported models
  • Multipliers only map model cost differences; billing stays consistent.
  • You do not need separate keys for separate models.
  • Quota does not expire or reset, which fits long-running testing and stable production usage.
Cache

Cache-Hit Billing

When part of a request matches previously processed content, the system can reuse that computation instead of charging the full input cost again.

What counts as a cache hitRepeated context can be reused

Cache reuse can improve response speed and reduce usage cost, especially in multi-turn chats, repeated workflow calls, and code-completion loops.

  • Normal input: billed at the model multiplier.
  • Repeated cached input: billed from 0.1x.
  • For high-cost models, cache multipliers may rise to 0.2x or be paused during extreme upstream volatility.
Simple example

If you repeatedly include a large shared context across a long conversation, the matched repeated portion can be charged at a lower multiplier.

Cache hits start at 0.1x

Matched cached input is billed at only 10% of the regular input cost.

Longer sessions can cost less

Follow-up chats, workflow calls, and code-completion loops are more likely to reuse repeated context.

Premium models may adjust

When upstream costs move sharply, cache multipliers may change and can be paused in extreme cases.

Note: cache hits depend on whether the request content qualifies for system cache reuse.

Reference

Model Price Reference

Compare multipliers, converted per-token cost, context size, and vision support before you buy.

claude-sonnet-4-6 concurrency route

claude-sonnet-4-6 is a concurrency-oriented route compatible with common client defaults, suitable for high-concurrency workflows, Python scripts, and text processing.

Premium model note

Premium model cost and availability may move with upstream supply. Use the live multiplier shown here and your actual call results as the final reference.

Python concurrency limit

Python-script concurrency is only allowed on claude-sonnet-4-6. Other models should not be used for high-concurrency scripts.

Updated from the latest model listCurrent converted prices and multipliers

A practical reference for understanding model cost before purchase.

Assumes max_tokens=8192
1x = CNY 40 / 100M tokens0.1x from CNY 0.04 / 1M tokens15x premium routes = CNY 6 / 1M tokensUp to 1M context
ProviderModelMultiplierExample: 2x means 100M tokens deducts 200M quota1M tokensConverted priceRecommended context windowAssumes max_tokens=8192Vision support
China aggregate route(High concurrency)
claude-sonnet-4-6Fast response
1xCNY 0.401M✅
MiniMax
MiniMax-M3
1xCNY 0.401M✅
MiniMax
image-01
per imageCNY 0.20
MiniMax
image-01-live
per secondCNY 2.00
StepFun
step-3.7-flashFast response
1xCNY 0.40256k✅
ByteDance
doubao-seed-2.0-code
2xCNY 0.80200k✅
ByteDance
doubao-seed-2.0-pro
2xCNY 0.80128k✅
美团 LongCat
LongCat-2.0Agentic Coding
4xCNY 1.601M
Xiaomi
mimo-v2.5-pro
1xCNY 0.401M✅
Xiaomi
mimo-v2.5
0.5xCNY 0.201M✅
DeepSeek
deepseek-v4-pro
7.5xCNY 3.001M
DeepSeek
deepseek-v4-flash
2.5xCNY 1.001M
Kimi
kimi-k2.6
2xCNY 0.80256K✅
Kimi
kimi-k2.7
4xCNY 1.60256K✅
Zhipu AI
glm-5.1
5xCNY 2.00256K✅
Zhipu AI
glm-5.2
8xCNY 3.201M✅
OpenAI
gpt-5.4
2xCNY 0.801M✅
OpenAI
gpt-5.5
4xCNY 1.60256k✅
OpenAI
gpt-5.3-codex-spark
4xCNY 1.60128k
OpenAI
gpt-image-2
per imageCNY 0.05-0.104K image output
Anthropic
claude-haiku-4-5-20251001
1xCNY 0.40256K✅
Anthropic
claude-sonnet-5高能力主力
12xCNY 4.801M✅
Anthropic
claude-fable-5顶级任务
100x40 元1M✅
Anthropic
claude-opus-4-6
15xCNY 6.001M✅
Anthropic
claude-opus-4-7
15xCNY 6.001M✅
Anthropic
claude-opus-4-8
15xCNY 6.001M✅
Google
gemini-3.1-pro
10xCNY 4.001M✅
Google
gemini-3.5-flash
6xCNY 2.401M✅
China aggregate routeContext 1M
(High concurrency)claude-sonnet-4-6Fast response
Multiplier1x
1M tokensCNY 0.40
context window1M
Vision✅
MiniMaxContext 1M
MiniMax-M3
Multiplier1x
1M tokensCNY 0.40
context window1M
Vision✅
MiniMax
image-01
Multiplierper image
1M tokensCNY 0.20
MiniMax
image-01-live
Multiplierper second
1M tokensCNY 2.00
StepFunContext 256k
step-3.7-flashFast response
Multiplier1x
1M tokensCNY 0.40
context window256k
Vision✅
ByteDanceContext 200k
doubao-seed-2.0-code
Multiplier2x
1M tokensCNY 0.80
context window200k
Vision✅
ByteDanceContext 128k
doubao-seed-2.0-pro
Multiplier2x
1M tokensCNY 0.80
context window128k
Vision✅
美团 LongCatContext 1M
LongCat-2.0Agentic Coding
Multiplier4x
1M tokensCNY 1.60
context window1M
XiaomiContext 1M
mimo-v2.5-pro
Multiplier1x
1M tokensCNY 0.40
context window1M
Vision✅
XiaomiContext 1M
mimo-v2.5
Multiplier0.5x
1M tokensCNY 0.20
context window1M
Vision✅
DeepSeekContext 1M
deepseek-v4-pro
Multiplier7.5x
1M tokensCNY 3.00
context window1M
DeepSeekContext 1M
deepseek-v4-flash
Multiplier2.5x
1M tokensCNY 1.00
context window1M
KimiContext 256K
kimi-k2.6
Multiplier2x
1M tokensCNY 0.80
context window256K
Vision✅
KimiContext 256K
kimi-k2.7
Multiplier4x
1M tokensCNY 1.60
context window256K
Vision✅
Zhipu AIContext 256K
glm-5.1
Multiplier5x
1M tokensCNY 2.00
context window256K
Vision✅
Zhipu AIContext 1M
glm-5.2
Multiplier8x
1M tokensCNY 3.20
context window1M
Vision✅
OpenAIContext 1M
gpt-5.4
Multiplier2x
1M tokensCNY 0.80
context window1M
Vision✅
OpenAIContext 256k
gpt-5.5
Multiplier4x
1M tokensCNY 1.60
context window256k
Vision✅
OpenAIContext 128k
gpt-5.3-codex-spark
Multiplier4x
1M tokensCNY 1.60
context window128k
OpenAIContext 4K image output
gpt-image-2
Multiplierper image
1M tokensCNY 0.05-0.10
context window4K image output
AnthropicContext 256K
claude-haiku-4-5-20251001
Multiplier1x
1M tokensCNY 0.40
context window256K
Vision✅
AnthropicContext 1M
claude-sonnet-5高能力主力
Multiplier12x
1M tokensCNY 4.80
context window1M
Vision✅
AnthropicContext 1M
claude-fable-5顶级任务
Multiplier100x
1M tokens40 元
context window1M
Vision✅
AnthropicContext 1M
claude-opus-4-6
Multiplier15x
1M tokensCNY 6.00
context window1M
Vision✅
AnthropicContext 1M
claude-opus-4-7
Multiplier15x
1M tokensCNY 6.00
context window1M
Vision✅
AnthropicContext 1M
claude-opus-4-8
Multiplier15x
1M tokensCNY 6.00
context window1M
Vision✅
GoogleContext 1M
gemini-3.1-pro
Multiplier10x
1M tokensCNY 4.00
context window1M
Vision✅
GoogleContext 1M
gemini-3.5-flash
Multiplier6x
1M tokensCNY 2.40
context window1M
Vision✅

Prices above are reference conversions based on current multipliers. Image models are billed per image or second. Actual usage follows the platform multiplier and request consumption; ✅ means vision is supported.

Next

Ready to buy after checking the price table

This page explains multipliers, cache hits, and model cost. New users can buy an API key; existing users can top up an existing key.

Get API accessRead integration docsTop up existing key