AI API Router Directory

Alibaba Cloud Model Studio (Bailian, 百炼) is Alibaba Cloud's managed model platform, hosting Qwen and third-party models across text, image, audio and video with pay-as-you-go billing per million tokens in CNY. The cn-beijing list rates on the official price page include qwen-long at 0.5 CNY per million input tokens and 2 CNY per million output tokens, qwq-plus at 1.6 CNY input and 4 CNY output, and deepseek-v3.2-exp at 2 CNY input and 3 CNY output, while international-deployment rates are materially higher (qwq-plus international: 5.871 CNY input, 17.614 CNY output). The price page states it shows list prices only ('原价') and that promotional pricing lives in the console, and pricing is tiered so a single request's token volume sets the unit price for that whole request.

Token allowance, not cash: first activation of Bailian grants a per-model free inference quota of typically 1,000,000 tokens (input and output combined, cn-beijing only, valid 90 days) per help.aliyun.com/zh/model-studio/new-free-quota; no USD-convertible value is published (checked 2026-09-16).

All API Hub is an open-source browser extension and account manager for the New-API/Sub2API ecosystem. It provides a unified dashboard for tracking balances, usage, and auto-checking across multiple relay providers, rather than serving as an API endpoint itself.

Free open-source tool that grants no model quota itself; its docs only teach users how to farm check-in/free allowances on other relays. No free API value is provided here.

Apishop is a professional AI API gateway designed for developers in mainland China. It provides direct, low-latency access to Anthropic, OpenAI, and Gemini models with zero code changes required. Billing is transparently mapped 1:1 to official USD rates but paid in local currency.

freeairouter review: register-then-pay-as-you-go commercial gateway, no public free quota seen; official site is a mainland-friendly gateway drop-in compatible with OpenAI/Anthropic/Gemini.

Baseten offers Model APIs, hosted open-weight models behind an OpenAI Chat Completions endpoint at inference.baseten.co with a beta Anthropic Messages endpoint, alongside dedicated GPU deployments and training billed per minute. Model API rates range from $0.10/1M input and $0.50/1M output for GPT OSS 120B to $3.00/1M input and $15.00/1M output for Kimi K3, with cached input billed at roughly a tenth of the input rate. The Basic plan is $0 per month pay-as-you-go, and the pricing FAQ states new accounts come with credits but publishes no amount.

The official pricing FAQ states that new Baseten accounts come with credits to experiment with deployments, but no dollar amount, expiry or conditions are published (checked 2026-09-16).

Black Forest Labs runs the hosted FLUX model API with a credit system where 1 credit equals USD $0.01, charged per image for stills and per second for video. FLUX 3 video costs $0.17/s at HD, $0.29/s FHD, $0.40/s QHD and $0.80/s UHD for text- and image-to-video, while video-to-video (continuation, editing, restyling) runs $0.41-$0.95/s; a draft tier renders HD at $0.06/s. FLUX.2 images start at $0.014 for [klein] 4B and rise to $0.07 for [max], with editing at $0.045 for [pro]. Clips are capped at 20 seconds and synchronized audio is included at no extra charge.

No free credit grant is stated on official pages; docs.bfl.ai/quick_start/get_started instructs users to add credits (suggesting $10-20 to start) with no signup bonus (checked 2026-09-16).

Chutes is a decentralized serverless AI inference marketplace on Bittensor (Subnet 64). It provides an OpenAI-compatible API for deploying and scaling open-source models using high-performance VLLM templates, with per-token pricing and optional monthly subscription plans.

Official announcement 2026-02-27: $5-deposit 200-requests/day plan closed to new users, fully retires 2026-03-15; now pay-per-token (or GPU by the second). Bittensor-subsidized cheap models are not a free tier.

Claude-ZH (claude-zh.cn) is a third-party relay aggregator optimized for Chinese developers, providing access to Claude models via a homegrown CLI ('lucky') and OpenAI-compatible endpoints with local recharge options.

Mainly a Claude Code Chinese tutorial/docs site; model calls require email signup and /recharge pay-as-you-go top-up; docs mention no free quota — only the tutorials are free.

CoderPlan is a commercial AI API relay focused on developer tools like Claude Code and Gemini CLI. It provides pay-as-you-go top-up tiers for Chinese developers, claiming to match official API pricing for models like Claude and GPT.

Small signup 'trial credit' whose amount is unstated — not quantified; no sustained free tier. Core is four paid top-up tiers $30/$100(+$5)/$300(+$20)/$500(+$50) with non-expiring credit.

Fuka API is a comprehensive AI API infrastructure provider offering a unified gateway to a vast selection of models via the New-API protocol. It supports high concurrency, transparent pay-as-you-go billing, and multi-protocol compatibility for developers and teams.

e7ai posts (2026-06/07): '芙芙卡' api.fuka.win gives signup credit ('注册就送100') + check-in; operator routerpark listing says '注册送150'. Amount/currency unverified; no free-priced models in /api/status.

GitHub Models was GitHub's hosted model playground, model catalog, inference API and bring-your-own-key service, and GitHub previously presented it as separate from GitHub Copilot. GitHub's official documentation now states that as of July 30, 2026, GitHub Models has been fully retired: the playground, model catalog, inference API and BYOK are no longer available to any customer. The same page directs users needing model access to Azure AI Foundry and users wanting AI workflows on GitHub to GitHub Copilot.

No free tier stated on official pages; GitHub Models was fully retired on July 30, 2026 with no playground, catalog, inference API or BYOK access remaining (checked 2026-09-16).

HvoyAI.com (formerly Hvoy.ai) is the primary domain for the Hvoy AI diagnostics platform. It provides API key testing tools, gateway rankings, and comparative performance data for AI relay providers across the global ecosystem.

No free LLM API of its own: /free-tokens page distributes third-party relay redemption codes/temp keys via points & daily lottery (~20 pts signup, ~10/redemption); quota, models, expiry set by third-party gateways.

Hyperbolic (hyperbolic.ai) is an AI cloud that sells on-demand, reserved and private GPU capacity and, separately, a serverless inference API described as OpenAI-compatible at base URL https://api.hyperbolic.xyz/v1 with 25+ open-source models. Its official rate-limit table defines a Basic tier of 60 requests per minute available to a free account with a $0 minimum deposit, and a Pro tier of 600 requests per minute that unlocks automatically once $5 or more is deposited. Serverless inference is pay-as-you-go: text generation runs from $0.10 per 1M tokens for small models, $0.20-$0.40 for 32B-72B models and $0.30-$4.00 for 120B-480B models, with images at a $0.01 base rate per 1024x1024 image at 25 steps and audio at $5.00 per 1M characters.

No free tier stated on official pages (checked 2026-09-16): Hyperbolic documents a free "Basic" tier as 60 requests/minute with a Free account, but no dollar credit or trial allowance is published on the pricing or billing pages.

Inference.net is a unified LLM gateway, observability and fine-tuning platform rather than an image or video vendor. Its pay-as-you-go tier is $0 + usage and includes 1M gateway requests and 1M tracing spans per month, 14-day data retention, one seat and a 30 req/min rate limit. Growth is $250/month with a $50 one-time opening credit, 50M gateway requests and tracing spans per month, unlimited seats and a 250 req/min limit. The catalog lists 60 models from OpenAI, Anthropic, Google and open-source labs billed per token, from Schematron V2 Turbo at $0.03/$0.15 to Claude Opus 5 and GPT-6 Astra at $5.00-$10.00 input and $25.00-$50.00 output per 1M tokens.

No free dollar credit: the $0 pay-as-you-go tier includes 1M gateway requests and 1M tracing spans per month, while the $50 one-time opening credit is listed only on the $250/month Growth plan (checked 2026-09-16).

Juzi AI (api.juziai.cc) is a Chinese commercial-aggregator relay built on the New-API framework. It aggregates high-end models like GPT-5 and Claude Opus by leveraging various upstream account pools, offering pay-as-you-go access to users in mainland China.

Free quota negligible: daily check-in after 10 calls grants 5–20 quota (quota_per_unit=500000 ≈ $1 per 500K quota) → well under $0.01/day; most model_ratio=0 models are pay-per-call. Not quantifiable.

LLM7.io is an open-source AI API gateway providing unified access to leading models like GPT and DeepSeek. It stands out for its unique anonymous access tier, allowing developers to test models without registration, and a registration-based free tier with higher limits.

Free relay: ~30 RPM anonymous, rising to ~120 RPM with free token from token.llm7.io (20/min, 100/h sub-caps); base URL api.llm7.io/v1 with api_key 'unused'. No token quota published → no USD estimate.

MakeLove NewAPI (Nianhua API) is an anonymously operated Chinese API relay on the New-API panel that aggregates various models with paid top-up packages as the primary business model. Details were checked against the provider's own page at newapi.makelove.cloud.

Free model entries remain (~13 per /api/pricing); affiliate posts advertise '$10 signup + daily check-in' but site announced removing new-user free credit; free models are the reliable free tier.

ModelScope (魔搭, modelscope.cn) is an open-source model and dataset community founded in June 2022 by the Institute for Intelligent Computing together with the CCF Open Source Development Committee, organised around 'Model-as-a-Service'. The platform publishes models, datasets, Studios (application display space), MCP and Skills listings, and maintains the ModelScope, Swift, EvalScope and ModelScope-Agent frameworks. Its stated free offering is platform credits: 'Sign up / Log in to get 200 Magicubes' plus 'Link Alibaba Cloud account, get 50 more daily', with Studios described as 'a free and flexible AI application display space'.

No cash value stated: the official home page says 'Sign up / Log in to get 200 Magicubes' and 'Link Alibaba Cloud account, get 50 more daily' (modelscope.cn, checked 2026-09-16); Magicubes have no published CNY or USD rate.

Modeloc (modeloc.com) is an evaluation and compute-sharing platform for LLM relays. It features a 'compute pool' where users can share spare model quota to earn credits or spend credits to call pooled models, integrated with a trust-score leaderboard.

Compute-pool swap: deposit your own account quota, billed by usage; guests get free model-consistency checks only. Referral grants both sides ~100K compute (~$10) one-off — not a public free API tier.

NVIDIA NIM's hosted API (build.nvidia.com / integrate.api.nvidia.com) serves a broad catalog of open-weight models — Nemotron, Llama, Qwen, DeepSeek, GLM and more — in an OpenAI-compatible shape, with a permanent free tier: API key at sign-up, no card, 40 requests/minute by default (200 RPM granted on request, per NVIDIA's developer forums).

Free tier is rate-limited (40 RPM default, 200 on request) rather than credit-based — no USD value to quantify.

Nebius sells two products relevant to an API-pricing decision: Token Factory, an OpenAI-compatible inference API for open models (where each model is offered in a default and a -fast flavor with different latency and token pricing), and GPU instances billed per GPU-hour, from $2.15 for preemptible H100 up to $7.85 on-demand for HGX B300. The public pricing page documents instance, storage and network rates plus billing rules; Token Factory per-token prices are only shown after login. Commitment reservations are advertised at up to 35% below on-demand rates.

No free tier or USD sign-up credit is stated on official pages (checked 2026-09-16); the docs describe promo codes that grant free credits only when Nebius issues one in a special offer, and the first card payment is $25 credited to the account balance.

NexusVAI is a multi-model API platform providing a unified, OpenAI-compatible endpoint for frontier LLMs. It features a developer-centric playground (CancriCode) and offers pay-as-you-go access to a broad range of open and proprietary models with an emphasis on ease of integration.

No public free-tier information found (researched 2026-08-28).

Novita AI runs 200+ models across text, image, audio and video behind an OpenAI-compatible API at api.novita.ai/openai, plus serverless endpoints, GPU instances and agent sandboxes, all billed per token or per GPU-second. Published model rates start at $0.02/1M input and $0.05/1M output for Llama 3.1 8B Instruct, with DeepSeek V4 Flash at $0.14/$0.28 on a 1M-token context, and three Ling 3.0 Flash variants (Fin, VL, Sante) listed at Free input and Free output with 256K context. Batch inference carries an introductory 50% discount on input and output tokens for supported models.

No standing free credit is stated on the current official pricing or docs pages (checked 2026-09-16); the pricing table lists Ling 3.0 Flash Fin/VL/Sante at Free input and output (256K context), and an official blog post from May 2025 advertised a limited-time $10 new-user credit that current pages do not repeat.

OVHcloud AI Endpoints is a serverless, OpenAI-compatible inference API for 40+ open-weight models, hosted in Gravelines, France, and billed per million tokens per model with input and output priced separately. Published catalogue rates include gpt-oss-20b at EUR 0.04 per 1M input and EUR 0.15 per 1M output, gpt-oss-120b at EUR 0.08/EUR 0.40, Meta-Llama-3.3-70B-Instruct at EUR 0.67/EUR 0.67 and Qwen3.5-397B-A17B at EUR 0.60/EUR 3.60, with embeddings at EUR 0.01 per 1M input tokens and whisper-large-v3 at EUR 0.00004083 per second of audio. Several models are listed as Free in the catalogue (Qwen3Guard-Gen-8B and Qwen3Guard-Gen-0.6B, stable-diffusion-xl-base-v10 and the nvr-tts voices), and a Batch tier in beta is documented at 50 percent savings for bulk processing.

No credit or allowance stated on official pages (checked 2026-09-16); the official AI Endpoints catalogue instead lists specific models as Free (Qwen3Guard-Gen-8B and -0.6B, stable-diffusion-xl-base-v10, and the nvr-tts-de-de/it-it/en-us/es-es voices) with all other models billed per million tokens.

Ollama (ollama.com) is primarily a local runtime for open models - the ollama.com pricing page lists a Free plan at $0 whose first feature is "Run models locally" and states plainly that "Local models are always free" - and it also operates a cloud API for models too large for local hardware. The cloud side is priced per million tokens with a published table (for example gpt-oss:20b at $0.07 input / $0.30 output, gemma4 at $0.14 / $0.40, deepseek-v4.1-flash at $0.15 / $0.60 and kimi-k3 at $3.00 / $15.00), with higher peak pricing between 12:00 and 18:00 UTC Monday to Friday. Paid plans bundle usage credits: Pro is $20/mo including $60 of usage, Max is $100/mo including $300, and Team is $500/mo including $1,000 of shared credits, while Free accounts receive a starter amount of usage on a smaller set of starter models and can add credits to unlock all models.

No free tier stated on official pages (checked 2026-09-16): the Ollama Free plan lists "Starter usage credits included" and "Includes access to starter models" at $0, but no dollar or token amount is published; local model usage is stated to be always free.