Platform
OpenRouter
An OpenAI-compatible multi-model routing entry point, useful for quickly comparing different providers and open models.
Global aggregationOne key, many modelsUSD pricing
View platform page β// APISpotlight Β· AI API Radar
Find AI APIs, check free quotas, and test an authorized endpoint directly
Discover official API platforms and developer resources, review free tiers and public verification status, then use a real request to see what an endpoint you are authorized to use actually returns.
Only test endpoints you are authorized to use; the key stays in this browser's memory only and is cleared once testing ends.
- API platforms
- 28
- Free tiers
- 16
- Dev resources
- 35
- Last benchmark
- 2026-09-29
Platform latency, success rate and status come from actual measured data fields; a test result describes that one observation and is not permanent certification.
CONTENT FOUNDATION
Start with the entry points that are actually useful
APISpotlight brings together reproducible benchmarks, standalone platform pages, verified free tiers and curated developer resources.
Online
Platform
OpenAI
The official platform offering general text, reasoning, embedding and multimodal APIs.
Official direct accessIndustry benchmarkGPT-5.6 family
View platform page βPlatform
Anthropic
Anthropic's official Claude API, emphasizing long context, tool use and a reliable developer interface.
Code-focusedClaude familyLong context
View platform page βPlatform
Google Gemini
The entry point to the Google Gemini API and AI Studio, providing multimodal models and Google ecosystem integration.
MultimodalVery long contextGemini family
View platform page βPlatform
Groq
A high-speed inference API for open models, offering an OpenAI-compatible developer experience.
Very fast inferenceLPU hardwareFree tier
View platform page βPlatform
DeepSeek
The official DeepSeek API, providing reasoning and general-purpose model endpoints compatible with the OpenAI SDK.
China-based open sourceExtremely cost-effective
View platform page βPlatform
SiliconFlow
A model API aggregation and inference platform aimed at Chinese-language developers, covering a range of open models.
China-based GPU cloudFull model lineupSignup credit
View platform page βPlatform
Together AI
A developer platform offering open model inference, fine-tuning and GPU workflows.
Full model poolOpen weightsAccelerated inference
View platform page βPlatform
Hugging Face
An open AI community platform connecting models, datasets, Spaces and Inference Providers.
Open-source ecosystemHuge model count
View platform page βPlatform
Cerebras
An inference API built on purpose-built AI hardware, offering low latency and support for a range of open models.
Ultra-fast inferenceWafer-scale chip
View platform page βVERIFIED FREE TIER
Free quotas and free tiers
Free Tier
SiliconFlow
Multi-model inference API
About Β₯14 / on the order of 20 million tokens (recorded benefit; not directly confirmed on the current official page)
Verified: 2026-08-28 Β· Medium confidence
View free conditions βFree Tier
OpenRouter
OpenAI-compatible multi-model routing API
25+ Free models; the Free plan applies only to free models (official Pricing)
Verified: 2026-08-28 Β· High confidence
View free conditions βFree Tier
Together AI
Open-model inference API
$1 initial credit (recorded benefit; not directly confirmed on the current official page)
Verified: 2026-08-28 Β· Medium confidence
View free conditions βFree Tier
Groq
High-speed open-model inference API
Free tier; fixed token allowance Unknown
Verified: 2026-08-28 Β· Medium confidence
View free conditions βFree Tier
Google Gemini
Gemini API
Official Pricing explicitly lists a Free Tier; input and output for the listed models can be used free of charge
Verified: 2026-08-28 Β· High confidence
View free conditions βFree Tier
NVIDIA NIM
API Catalog model inference plus NIM self-hosting
The NVIDIA Developer Program provides free tools and learning resources for developers; the API Catalog explicitly offers Free serverless APIs for development; fixed API credits/allowance are not published
Verified: 2026-08-29 Β· Medium confidence
View free conditions βDEVELOPER RESOURCES
Developer resources
Free AI APIs28/30
Free AI APIs starter list
A collection of entry points built around verifiable free tiers, starting credits and low-friction APIs.
The directory is free to browse; each item's conditions must be checked against the official source.
View resource βFree AI Models28/30
Hugging Face Models
The official directory for open models, model cards, licenses and inference entry points.
The model directory is free to browse; whether inference is free depends on the provider and account.
View resource βFree API Credits27/30
Free API Credits
New-user credits, trial balances and promotional vouchers compiled from official sources.
The page is free; credit status must be confirmed on the provider's official page.
View resource βDeveloper Tools29/30
LiteLLM
Unifies multiple LLM providers behind an OpenAI-compatible interface, suited to gateways, routing and cost observability.
Open-source project, self-hostable; cloud service plans must be checked separately.
View resource βDeveloper Tools27/30
LangChain
A development framework for orchestrating models, tools, retrieval and agent workflows.
The core code is open source; third-party model/API costs are not included in the project.
View resource βDeveloper Tools26/30
LlamaIndex
An LLM application framework for RAG, data connectors and agents.
The open-source core is self-hostable; model, vector store and managed service costs are billed separately.
View resource β01 Β· BENCHMARK
π Measured official APIs
Click a column header to sort Β· latest measured data Β· source src/data/platforms.json
| # | Status | Authenticity probe | π¨π³ Reachable in China | Models | Entry | |||
|---|---|---|---|---|---|---|---|---|
| 1 | OpenRouterπ’ Officialβ Featured Global aggregationOne key, many modelsUSD pricing | Online | Direct International credit card Β· Cryptocurrency | 38 | 100.0% | openai/gpt-5.6-luna Β· anthropic/claude-opus-5 Β· google/gemini-3.7-flash +1 | Website β | |
| 2 | Cerebrasπ’ Official Ultra-fast inferenceWafer-scale chip | Online | Direct International credit card | 57 | 100.0% | llama-3.3-70b Β· llama3.1-8b Β· qwen3-32b | Website β | |
| 3 | Fireworks AIπ’ Official Accelerated inferenceOpen weightsEnterprise-grade | Online | Direct International credit card | 69 | 100.0% | accounts/fireworks/models/llama-v3p1-70b-instruct Β· accounts/fireworks/models/deepseek-v3 Β· accounts/fireworks/models/qwen2p5-72b-instruct | Website β | |
| 4 | Google Geminiπ’ Official MultimodalVery long contextGemini family | Online | Direct International credit card | 81 | 100.0% | gemini-3.7-flash Β· gemini-3.1-pro Β· gemini-2.5-pro | Website β | |
| 5 | Anthropicπ’ Official Code-focusedClaude familyLong context | Online | Direct International credit card | 88 | 100.0% | claude-opus-5 Β· claude-sonnet-5 Β· claude-haiku-4-5 | Website β | |
| 6 | xAI (Grok)π’ Official Grok familyReal-time data | Online | Direct International credit card | 94 | 100.0% | grok-4.6 Β· grok-4.5 Β· grok-4.3 | Website β | |
| 7 | Perplexityπ’ Officialβ Featured Web searchReal-time citations | Online | Direct International credit card | 98 | 100.0% | sonar Β· sonar-pro Β· sonar-reasoning | Website β | |
| 8 | Hugging Faceπ’ Officialβ Featured Open-source ecosystemHuge model count | Online | Direct International credit card | 109 | 100.0% | meta-llama/Llama-3.3-70B-Instruct Β· deepseek-ai/DeepSeek-V4-Pro Β· Qwen/Qwen3.7-Max | Website β | |
| 9 | Together AIπ’ Officialβ Featured Full model poolOpen weightsAccelerated inference | Online | Direct International credit card | 111 | 100.0% | deepseek-ai/DeepSeek-V4-Pro Β· meta-llama/Llama-3.3-70B-Instruct Β· Qwen/Qwen3.7-Max | Website β | |
| 10 | Groqπ’ Official Very fast inferenceLPU hardwareFree tier | Online | Direct International credit card | 133 | 100.0% | llama-3.3-70b-versatile Β· gemma2-27b-it Β· mixtral-8x7b-32768 | Website β | |
| 11 | OpenAIπ’ Official Official direct accessIndustry benchmarkGPT-5.6 family | Online | Direct International credit card | 169 | 100.0% | gpt-5.6-sol Β· gpt-5.6-terra Β· gpt-5.6-luna | Website β | |
| 12 | Mistral AIπ’ Official Europe-basedOpen-source modelsMultilingual | Online | Direct International credit card | 179 | 100.0% | mistral-large-latest Β· mistral-small-latest Β· codestral-latest | Website β | |
| 13 | Cohereπ’ Official Enterprise-gradeRAG ecosystemMultilingual | Online | Direct International credit card | 191 | 100.0% | command-r-plus Β· command-r Β· command-a | Website β | |
| 14 | Novita AIπ’ Official Cost-effectiveGPU cloud | Online | Direct International credit card | 274 | 100.0% | deepseek/deepseek-v4-pro Β· qwen/qwen3.7-max Β· meta-llama/llama-4-maverick-17b-128e-instruct-fp8 | Website β | |
| 15 | Zhipu AI (GLM)π’ Official Top tier in ChinaGLM family | Online | Direct Alipay Β· WeChat Pay | 291 | 100.0% | glm-4-plus Β· glm-4.5 Β· glm-4-air | Website β | |
| 16 | Agnes AIπ’ Official Omni-modalFree tier | Online | Direct Payment methods: refer to the official page | 307 | 100.0% | agnes-1.5-flash Β· agnes-2.0-flash Β· agnes-2.5-flash +3 | Website β | |
| 17 | DeepInfraπ’ Official Low-cost open weightsElastic | Online | Direct International credit card | 336 | 100.0% | deepseek-ai/DeepSeek-V4-Pro Β· anthropic/claude-opus-5 Β· meta-llama/Llama-3.3-70B-Instruct | Website β | |
| 18 | SambaNovaπ’ Official Wafer-scale chipOfficial R1Very fast | Online | Restricted International credit card | 349 | 100.0% | Meta-Llama-3.3-70B-Instruct Β· DeepSeek-R1-Distill-Llama-70B Β· Qwen2.5-72B-Instruct | Website β | |
| 19 | DeepSeekπ’ Officialβ Featured China-based open sourceExtremely cost-effective | Online | Direct Alipay Β· WeChat Pay | 410 | 100.0% | deepseek-v4-pro Β· deepseek-v4-flash | Website β | |
| 20 | StepFunπ’ Official China-based, fast-movingStep family | Online | Direct Alipay Β· WeChat Pay | 718 | 100.0% | step-2-16k Β· step-1-32k Β· step-1-8k | Website β | |
| 21 | Volcengine Ark (Doubao)π’ Official ByteDance ecosystemDoubao familyEnterprise-grade | Online | Direct Alipay Β· WeChat Pay +1 | 775 | 100.0% | doubao-pro-128k Β· doubao-pro-32k Β· doubao-lite-32k | Website β | |
| 22 | Alibaba Cloud Model Studio (Tongyi)π’ Official Alibaba ecosystemQwen modelsEnterprise-grade | Online | Direct Alipay Β· WeChat Pay +1 | 778 | 100.0% | qwen-max Β· qwen-turbo Β· qwen-plus | Website β | |
| 23 | SiliconFlowπ’ Officialβ Featured China-based GPU cloudFull model lineupSignup credit | Online | Direct Alipay Β· WeChat Pay | 784 | 100.0% | deepseek-ai/DeepSeek-V4-Pro Β· Qwen/Qwen3.7-Max Β· THUDM/GLM-4-9B-Chat | Website β | |
| 24 | MiniMaxπ’ Official Long contextChina-based startup | Online | Direct Alipay Β· WeChat Pay | 1103 | 100.0% | abab6.5s-chat Β· MiniMax-M1 | Website β | |
| 25 | Moonshot AI (Kimi)π’ Officialβ Featured Very long contextKimi familyChina-based, fast-moving | Online | Direct Alipay Β· WeChat Pay | 1239 | 100.0% | moonshot-v1-8k Β· moonshot-v1-32k Β· kimi-k2 | Website β | |
| 26 | Baidu Qianfan (ERNIE)π’ Official Baidu ecosystemERNIEEnterprise-grade | Online | Direct Alipay Β· WeChat Pay +1 | 1419 | 100.0% | ernie-4.0-8k Β· ernie-4.5 Β· ernie-3.5-8k | Website β | |
| 27 | Tencent Hunyuanπ’ Official Tencent ecosystemHunyuan family | Online | Direct WeChat Pay Β· Corporate bank transfer | 1779 | 100.0% | hunyuan-turbo Β· hunyuan-pro Β· hunyuan-large | Website β | |
| 28 | NVIDIA NIMπ’ Official NVIDIA ecosystemGPU cloudHigh throughput | Offline | Restricted International credit card | β | 0.0% | meta/llama-3.3-70b-instruct Β· deepseek-ai/deepseek-r1 Β· moonshotai/kimi-k2 | Website β |
How to read these numbers: latency is the median of the latest probe (3 samples per site); success rate = successful requests / total requests. Data is refreshed periodically by scripts/ping_test.py, and clicking a column header sorts the table. The authenticity column shows Phase 3 probe results (scripts/authenticity_test.py); rows show β / β³ when no test key is configured or the fingerprint baseline is not calibrated. Click a badge to expand probe details (temperature set / self-reported ID / drift / response / timestamp). The π¨π³ reachable-in-China column is what this site observed from a single machine and is not an official guarantee, since cross-border networks vary by time and route. Payment methods are compiled from public information; refer to the official page. The model list was compiled from official documentation and public model lists as of 2026-08; refer to the official page.
π Official public free credits and perks
New-user benefits published by each provider Β· refer to the official page Β· source src/data/perks.json
Trial credits on sign-up
SiliconFlow
New users receive trial credits after registering and completing verification (approximately Β₯14, on the order of 20 million tokens), usable with models across the whole platform; the exact amount is subject to the current announcement on the official site.
Condition: New user registration + verification
Valid until: Ongoing (refer to the official page)
Last verified: 2026-08-26
Claim β$10 trial credit on sign-up
OpenRouter
New users receive $10 in initial credits on registration, which can be used to test the aggregated full model pool; the specific policy is subject to the current announcement on the official site.
Condition: New user registration
Valid until: Ongoing (refer to the official page)
Last verified: 2026-08-26
Claim β$1 initial credit on sign-up
Together AI
New users receive $1 in free credits on registration, supporting testing across the full model pool, with rate limits on some models; refer to the official site.
Condition: New user registration
Valid until: Ongoing (refer to the official page)
Last verified: 2026-08-26
Claim βFree tier, no credit card required
Groq
Developers can use the free tier directly to test mainstream open models, with per-model RPM/TPM rate limits; refer to the official site.
Condition: Free registration
Valid until: Ongoing
Last verified: 2026-08-26
Claim βFree serverless APIs for development
NVIDIA NIM
The NVIDIA Developer Program provides free tools and learning resources for developers; the API Catalog publicly states that it provides Free serverless APIs for development. Fixed quotas, model coverage and production eligibility are subject to the account and terms; this item is not to be described as permanently free or free for production.
Condition: NVIDIA account + Get API Key
Valid until: Subject to the official policy
Last verified: 2026-08-29
Claim βMonthly free quota with a trial key
Cohere
Registration provides a trial API key with a monthly free call quota (on the order of a thousand calls); refer to the official site.
Condition: Free registration for a trial key
Valid until: Ongoing (refer to the official page)
Last verified: 2026-08-26
Claim βFree model inference quota + PRO subscription
Hugging Face
Inference Providers offers a rate-limited trial quota for free models; heavy individual use can subscribe to PRO (from about $9/month); refer to the official site.
Condition: Free registration (PRO requires a subscription)
Valid until: Ongoing (refer to the official page)
Last verified: 2026-08-26
Claim βNew-user sign-up credits
Zhipu AI (GLM)
After new users register at open.bigmodel.cn, trial credits are issued according to the official promotion and can be used with the full GLM model family; refer to the official site.
Condition: New user registration
Valid until: Subject to the official promotion
Last verified: 2026-08-26
Claim βNew-user free credits and free call volume
Alibaba Cloud Bailian (Tongyi)
The Bailian platform provides trial credits for new users, and some models offer a daily free token quota; refer to the official site.
Condition: New user registration (real-name verification required)
Valid until: Subject to the official promotion
Last verified: 2026-08-26
Claim βNew-user free resource package / voucher
Baidu Qianfan (ERNIE)
The Qianfan platform issues free resource packages/vouchers to new users for model trials, and the promotion runs according to the official announcement; refer to the official site.
Condition: New user registration (real-name verification required)
Valid until: Subject to the official promotion
Last verified: 2026-08-26
Claim βFree API pricing
Agnes AI
Official documentation shows that agnes-2.0-flash input/output tokens are currently $0 / 1M (standard price $0.03 / $0.15); the official site states that it provides free AI API credits and a free API key, which can be created on registration. Specifics are subject to the official site.
Condition: Register and create an API key
Valid until: Ongoing (refer to the official page)
Last verified: 2026-08-27
Claim βWorkers AI daily free Neurons
Cloudflare Workers AI
The official Pricing page states a free allocation of 10,000 Neurons per day, reset daily at UTC; some models require a paid billing method, and specific plan and model conditions are subject to the official site.
Condition: Cloudflare account + Workers AI
Valid until: Per Workers plan policy (refer to the official page)
Last verified: 2026-08-29
Claim βServerless Inference $1 Free Credits
Fireworks AI
The official Pricing page states that new users starting with Serverless Inference receive $1 in free credits; this is a one-time trial credit, not a permanently free API.
Condition: Register a Fireworks account
Valid until: Fixed validity period Unknown; subject to account policy
Last verified: 2026-08-29
Claim βOpen-weight self-hosting (not a free hosted API)
Mistral AI
Mistral's official pricing FAQ states that some open-weight models can be self-hosted; Mistral 7B and others are Apache 2.0 for research/individual use, and commercial deployment requires checking the license separately.
Condition: Own GPU/infrastructure + use under the model license
Valid until: Subject to the model license and official terms
Last verified: 2026-08-29
Claim β