小天公益站 provides a unified API gateway for large language models, enabling developers to access various AI models through a single interface. Key endpoints include /v1/chat/completions, /v1/embeddings, /v1/images/generations, /v1/audio/transcriptions, and /v1beta/models. It supports models from providers such as MoonshotAI, OpenAI, Grok, Zhipu, Volcengine, Cohere, Claude, Gemini, Suno, Minimax, Wenxin, Spark, Qingyan, DeepSeek, Qwen, Midjourney, AzureAI, Hunyuan, and Xinference. The service emphasizes better pricing and stability, with features like rate limiting (e.g., 25 requests per 3 minutes for default groups, 150 requests per 3 minutes for translation groups) and model grouping for optimized use cases like translation. New users receive an initial credit of 5 yuan upon registration.

Model

Input ($/M)

Output ($/M)

Audit

Speed

Latency

glm-5.2nvidia

$1.00/M

$4.40/M

—

MiniMax-M3nvidia

$0.600/M

$2.40/M

—

step-3.7-flashnvidia

$0.400/M

$2.30/M

—

deepseek-v4-pronvidia

$0.696/M

$1.39/M

—

deepseek-v4-flashnvidia

$0.224/M

$0.448/M

—

MiniMax-M2.7nvidia

$0.600/M

$2.40/M

—

Translation

—

242.5 t/s

2.56 s

step-3.5-flashnvidia

$0.200/M

$0.600/M

—

glm-4.7-flashdefault

$0.080/M

$0.600/M

—

GLM-4.7

—

23.5 t/s

23.07 s

gemini-2.5-flash-litedefault

$0.180/M

$0.720/M

—

gpt-oss-20bnvidia

$0.080/M

$0.300/M

—

glm-4.5-air

—

78.3 t/s

18.76 s

Time

Model

Speed

Latency

Apr 3, 03:53 PM

glm-4.5-air

78.35 tok/s

18.76s

Apr 3, 03:52 PM

Translation

203.11 tok/s

2.56s

Apr 3, 03:52 PM

GLM-4.7

23.53 tok/s

23.07s

Mar 8, 10:00 AM

Translation

281.81 tok/s

2.57s

Provider

Why compare

Models

Free

Avg price

Speed

30d uptime

小天公益站

new-api-xt-url-com

A unified gateway API for large language models, offering access to multiple providers with competitive pricing and stability.

Current provider baseline

110

$851.43/M

150 tok/s

99.6%

ai-chuyel-top

A free, community-supported API service providing access to various AI models for developers and users.

Lower average pricing
Faster measured speed
More free-model options
Broader model coverage
Same provider category

237

187

$0.646/M

edge-pieixan-icu

Pie-Xian (小水管) provides an OpenAI-compatible API relay with multiple endpoints including api.pie-xian.com and edge nodes.

Lower average pricing
Faster measured speed
More free-model options
Broader model coverage
Same provider category

175

$9.30/M

ai-cuz-lab-space

Cuz AI runs an OpenAI-compatible relay at ai.cuz-lab.space with broad model coverage, public pricing, and stable throughput for chat and coding workloads.

Lower average pricing
Faster measured speed
Broader model coverage
Same provider category

143

$62.07/M

280 tok/s

46.6%

api-6345ywz-cn

6345ywz API is an OpenAI-compatible API relay providing access to multiple AI models with competitive pricing.

Lower average pricing
Faster measured speed
Higher 30-day availability
Same provider category

105

$0.335/M

241 tok/s

99.8%

huashang-dpdns-org

An OpenAI-compatible API relay providing access to multiple AI models with extensive model coverage and pricing data.

More free-model options
Broader model coverage
Same provider category

266

$19625489189.65/M

57 tok/s

98.4%

ai-071129-xyz

MapleLeaf API runs a New API-powered gateway on ai.071129.xyz for aggregated access to multiple AI models.

More free-model options
Broader model coverage
Same provider category

263

292

N/A

38 tok/s

99.3%

Announcements

success4/29/2026

The default group deepseek-v4 series models on this site are all reverse-engineered from the official site. Tool calls may have issues, and token speed is relatively low. If you mind, do not use.

Additionally: The Guangzhou Tencent Cloud line has been taken offline due to machine expiration. Please switch to other lines.

default4/4/2026

Due to codex registration risk control issues, we are unable to provide the codex channel in large quantities. Therefore, this channel will be taken offline.

ongoing3/17/2026

LDC is temporarily closed, so recharging is currently unavailable. If needed, you can send a private message to obtain a redemption code.

success3/12/2026

Due to high memory usage by CPA, we have switched the reverse proxy scheme for codex. Theoretically, this switch is seamless. If issues arise, please contact via private message.

success3/12/2026

Launched the codex unlimited package, limited to 3 purchases per day. Allows unlimited use of codex channel models (group multiplier 0).

Currently in testing, the temporary price is 100, and it will increase later.

success2/14/2026

Some adjustments have been made to the flashx models in the official account pool. Details are as follows: glm-4-flashx-250414 -> glm-4-flash-250414 glm-4.7-flashx -> glm-4.7-flash

To ensure model name compatibility, flashx will be automatically routed to the same flash model.

flashx is actually a flash model with higher output speed. This adjustment does not affect model output quality.

success1/27/2026

Added the Guangzhou Tencent Cloud line. Do not use GPT models on this line. Uses cluster deployment, and the machines do not have external network access.

success1/13/2026

The z-image-turbo model is now available. Based on 2api, it does not support NSFW. Friends with this need can look for alternatives. Important reminder: Call via the drawing interface /v1/images/generations, not the chat interface. It has been tested and works in Cherry Studio.

success1/7/2026

Now integrated with Linux Do Credit. The original method of obtaining redemption codes via private message will continue to be available.

default12/23/2025

About Balance

New users receive 5 yuan upon registration. When used up, contact via private message to obtain a redemption code. No check-in, no AFF.

default12/23/2025

About Rate Limiting

This is the first time running a public welfare site, and I'm not sure how to configure reasonable values. The current limit is set to a rate that doesn't easily trigger restrictions in my own usage scenario. If friends have better configuration suggestions, feel free to suggest.

default12/23/2025

About Available Models

Currently, only the Translation and default groups are available. Other models are for personal use. Currently, only Zhipu-related models are provided (via account pool polling, not 2api).

About Groups

The default group is a general group with no usage scenario restrictions, but NSFW scenarios may trigger Zhipu's own review, leading to unavailability. Rate limit: 25 times per 3 minutes.

The Translation group is dedicated to translation. To ensure translation speed, models that support setting thinking have thinking unifiedly disabled. Rate limit: 150 times per 3 minutes.

Notes

Health checks: Scope: the 72-hour chart and recent availability measure API connectivity only. Each bar summarizes one hour of checks. Targets: LMSpeed tries the configured health check URL and provider status URL first, then API endpoints derived from known API hosts and recent speed-test base URLs. A website host is considered only when it looks like an API endpoint. Probe steps: each candidate goes through DNS lookup, TCP connection, TLS handshake for HTTPS, and an HTTP HEAD request with redirects followed. Probing stops after the first reachable candidate. Reachable criteria: every required network step must succeed. An HTTP response below 500 is treated as reachable, including 401 because it confirms that an authenticated API endpoint responded, except for statuses classified as blocked. Blocked results: HTTP 403, 429, 521, 525, and 530, plus detected WAF or Cloudflare challenges, are shown as blocked and excluded from availability calculations because LMSpeed cannot determine whether the API itself is down. Model availability: when a dedicated test key is configured, LMSpeed sends an authenticated GET request to a derived /models endpoint and compares returned model IDs with this provider's listed models. These per-model results appear in Models & Pricing and are not included in the provider connectivity percentage. Timeouts: TCP connection, TLS handshake, HTTP connectivity, and model requests each use a 5-second timeout. A full run can take longer when several candidates are tried. Frequency: a background worker checks all providers every 5 minutes by default. The 72-hour chart combines those samples into hourly bars, and the schedule may be changed by the service operator. Limit: automated samples are not an SLA and do not guarantee account quota, every model, every region, or successful completion requests. Check the provider's own status page before making operational decisions.

Domain Rating data is sourced from Ahrefs. It is a 0–100 backlink-based domain strength signal and does not measure API speed or reliability.

Announcements and FAQ are read from this provider's NewAPI status snapshot when available. LMSpeed stores the original content and optional English translations from the provider status source, then shows the localized fields on this page.

小天公益站

小天公益站

Features

Login Methods

API Endpoints

Leaderboard Rankings

Health Check

API Benchmarks & Pricing

Recent Test Records

Similar API Provider Alternatives to Compare

Announcements

About Balance

About Rate Limiting

About Available Models

About Groups

The default group is a general group with no usage scenario restrictions, but NSFW scenarios may trigger Zhipu's own review, leading to unavailability. Rate limit: 25 times per 3 minutes.

Notes

Similar API Provider Alternatives to Compare

Leaderboard Rankings

Provider	Why compare	Models	Free	Avg price	Speed	30d uptime
小天公益站 new-api-xt-url-com A unified gateway API for large language models, offering access to multiple providers with competitive pricing and stability.	Current provider baseline	110	2	$851.43/M	150 tok/s	99.6%
ai-chuyel-top A free, community-supported API service providing access to various AI models for developers and users.	Lower average pricing Faster measured speed More free-model options Broader model coverage Same provider category	237	187	$0.646/M
edge-pieixan-icu Pie-Xian (小水管) provides an OpenAI-compatible API relay with multiple endpoints including api.pie-xian.com and edge nodes.	Lower average pricing Faster measured speed More free-model options Broader model coverage Same provider category	175	10	$9.30/M
ai-cuz-lab-space Cuz AI runs an OpenAI-compatible relay at ai.cuz-lab.space with broad model coverage, public pricing, and stable throughput for chat and coding workloads.	Lower average pricing Faster measured speed Broader model coverage Same provider category	143	0	$62.07/M	280 tok/s	46.6%
api-6345ywz-cn 6345ywz API is an OpenAI-compatible API relay providing access to multiple AI models with competitive pricing.	Lower average pricing Faster measured speed Higher 30-day availability Same provider category	105	0	$0.335/M	241 tok/s	99.8%
huashang-dpdns-org An OpenAI-compatible API relay providing access to multiple AI models with extensive model coverage and pricing data.	More free-model options Broader model coverage Same provider category	266	43	$19625489189.65/M	57 tok/s	98.4%
ai-071129-xyz MapleLeaf API runs a New API-powered gateway on ai.071129.xyz for aggregated access to multiple AI models.	More free-model options Broader model coverage Same provider category	263	292	N/A	38 tok/s	99.3%

小天公益站

小天公益站

Features

Login Methods

API Endpoints

Leaderboard Rankings

About 小天公益站

Health Check

API Benchmarks & Pricing

Recent Test Records

Similar API Provider Alternatives to Compare

Announcements

About Balance

About Rate Limiting

About Available Models

About Groups

The default group is a general group with no usage scenario restrictions, but NSFW scenarios may trigger Zhipu's own review, leading to unavailability. Rate limit: 25 times per 3 minutes.

Notes

Similar API Provider Alternatives to Compare

Leaderboard Rankings