AI API Router Directory

Cerebras Inference offers ultra-fast AI inference powered by its Wafer-Scale Engine (WSE-3) technology, achieving record-breaking speeds up to 3000 tokens/s. The official API is OpenAI-compatible and features a generous free tier alongside professional pay-as-you-go access.

Free tier ≈1M tokens/day (resets UTC 00:00, no card, no waitlist) → ≈$30/mo blended at $1/1M; 8K context cap; RPM reported 5–30 by different sources. Paid dev tier from $10.

Lightning AI (LitAI) provides a unified gateway to frontier models through an OpenAI-compatible API. New users receive 40 million free tokens (approx. $15 in credits) to test models like Claude, GPT, and Gemini. Usage beyond the initial credit is billed as pay-as-you-go.

~30M free tokens per user per month (~15 credits = $15 value), no card required; credits refresh monthly and expire if unused; plus 1 free 4-CPU Studio and 10GB Drive.

NLP Cloud is an API platform offering pre-trained and custom models for summarization, classification, embeddings, ASR, speech synthesis and large language models such as GPT-OSS 120B, LLaMA 3.1 405B and ChatDolphin. Its pay-as-you-go plan has no fixed monthly cost and automatically grants a $15 free credit, then bills per request at $0.003 on CPU or $0.005 on GPU with token surcharges for the largest models. Published prepaid plans run from Starter at $29 per month (10 parallel requests) to the Large Language Models plan at $2,499 per month (50 parallel requests), with dedicated fine-tuning plans from $399 per month.

Official pricing page states that the pay-as-you-go plan automatically gives a $15 FREE credit with no fixed monthly cost (nlpcloud.com/pricing.html, checked 2026-09-16).

Helyx markets itself as a unified OpenAI-compatible API gateway for frontier models (Claude Opus 5, Claude Haiku 4.5, GPT-5.4 Nano, Gemini, Kimi, DeepSeek) claiming 50+ models behind one endpoint, with a published rate card of five active models including DeepSeek-V4 Pro at $0.400 in / $0.750 out per 1M tokens (50% off its $0.800/$1.50 list) and MiniMax M3 at $0.300/$1.200. Signup is Google or GitHub with no password, and the account is credited with $10.000000 free: $2.000000 instantly plus $8.000000 released on request. Paid top-ups are crypto-only (USDT, USDC, BTC, ETH, TRON) and every deposit is doubled ($5 pays $10 credit, $25 pays $50), with credits stated to never expire and no subscriptions or daily caps. The domain in this listing, helyx.ai, is dead - it returns a 404 DEPLOYMENT_NOT_FOUND from Vercel - and the live Helyx AI site is served from helyxai.space.

$10.000000 signup credit ($2.000000 instant, $8.000000 released on request) as stated on the helyxai.space homepage and pricing page; the $10 figure is the operator's own published grant, not a conversion.

Cerebras Inference serves open-weight models on wafer-scale CS-4 hardware — gpt-oss-120b at ~3,000 tokens/s for $0.35/$0.75 per 1M in/out, Qwen 3.8 27B at $0.99/$1.49. The self-serve developer tier is pay-as-you-go with $5 of free credit to start; the enterprise tier (production capacity, priority limits, fine-tuning) is quote-based. Also available via AWS Marketplace, OpenRouter, Hugging Face and Vercel.

Self-serve developer tier includes $5 free credit to start (pricing page, checked 2026-09-16). Replaces an older $30 figure that no longer matches the official page.

Nscale's Serverless Inference is an OpenAI-compatible endpoint service for open models - the catalogue pages list Meta Llama 4 Scout, Alibaba Qwen, DeepSeek, OpenAI GPT OSS 20B/120B, Mistral 8x22B Instruct, FLUX.1 [schnell] and Stable Diffusion XL Base 1.0 - with token-based pay-as-you-go billing. Per Nscale's own launch materials, pricing is based on input and output tokens for text and vision models, while image generation is priced on output image dimensions. Per-token rates are not published on a static official page: the docs direct users to the console's AI Services to Models list or the /v1/models API endpoint for pricing and context length. Every new user can claim $5 of free credit, and Nscale states the platform does not log or train on request or response content.

Every new user can claim $5 of free credit, stated in Nscale's Serverless Inference launch post (2 April 2025) and in its announcement that pricing is based on input and output tokens for text and vision models.

AnyAPI is a commercial multi-model API aggregation gateway providing access to 400+ models via a single key. It offers a generous free tier for developers and scales to high-volume enterprise plans, supporting models from OpenAI, Anthropic, and open-weights families.

Free tier $0/mo: 100K tokens/day (~100 credits/day, 1 credit≈1K tokens), ~20 RPM / 200 RPD, free and basic models only, no card. ≈$3/mo blended at $1 per 1M tokens.

OpenRouter is a unified, OpenAI-compatible API aggregating hundreds of model providers behind one key. Free model variants (:free IDs) are genuinely free with documented platform caps — 20 requests/min and 50/day for accounts that have purchased under 10 credits, 1,000/day once you've purchased 10+ — while paid variants have no platform request cap. Fallback routing, per-key credit caps, and BYOK are built in.

:free models cost $0, no card; 20 req/min; 50 :free req/day with <10 credits, 1000/day after buying $10+ credits (verified FAQ; failed reqs count). ~$1.5/mo equivalent at 50 req/day, ~1K tokens/call.

Venice (venice.ai) — privacy-focused web app (anonymous or zero-retention) for chat, image, video and audio, known for self-hosted open-weight models plus TEE-hosted closed models. Unofficial browser/cookie integration (merged into existing 'venice' row).

Credits are USD-denominated; Pro $18/mo → $1 credit, Pro+ $68/mo → $75 credit (10% bonus), Max $200/mo → $225 credit (12.5% bonus).

The Cloudflare AI Playground at playground.ai.cloudflare.com is a browser environment for experimenting with, benchmarking and tuning AI models, with a Model mode for testing system prompts and run settings, a Compare mode for side-by-side evaluation of models under identical prompts, and an experimental Generative UI mode. It is the front end for models that Cloudflare hosts and serves, and the free access terms come from Workers AI: the free allocation allows anyone to use a total of 10,000 Neurons per day at no charge, with additional usage on the Workers Paid plan billed at $0.011 per 1,000 Neurons. The model catalog covers 65 models including Llama, Qwen, DeepSeek, Kimi, GLM, Gemma, Mistral, Nemotron, Whisper and FLUX families, and the playground itself states that output is unverified and does not reflect Cloudflare's views.

Free allocation is stated as 10,000 Neurons per day at no charge on the Workers AI pricing page, equivalent to about $0.11/day at the published rate of $0.011 per 1,000 Neurons (the dollar figure is a conversion, the allowance is official).

SambaNova Cloud serves open-weight models (DeepSeek-V3.1, DeepSeek-V3.2, Meta-Llama-3.3-70B-Instruct, gpt-oss-120b, gemma-4-31B-it, MiniMax-M2.7 and MiniMax-M3) through an OpenAI-compatible endpoint at api.sambanova.ai/v1, priced per 1M tokens from $0.22 input / $0.59 output for gpt-oss-120b up to $3 / $4.50 for DeepSeek-V3.1 and V3.2. The free tier is defined by the absence of a payment method: 20 requests per minute, 20 requests per day and 200,000 tokens per day per model, on the five models listed in the free-tier tables. Linking a card moves the account to the Developer tier, which raises per-model limits and caps the account at 20M tokens per day across all models.

Free tier is a recurring daily allowance, not a USD credit: 200,000 tokens/day (20 RPM, 20 RPD) per the official rate-limits docs; 200k tokens at the cheapest published input rate of $0.22/1M (gpt-oss-120b) is about $0.04/day, while the same 200k at DeepSeek-V3.1's $3/1M input rate would be about $0.60/day.

Antigravity CLI (agy) — the terminal TUI for Google Antigravity 2.0, replacing Gemini CLI. Shares the same agent backend as the Antigravity IDE (incl. Claude models). Authenticates via Google OAuth. Free 'Individual' tier with weekly rate limits; higher limits via Google AI Pro/Ultra plans. No bring-your-own-key support.

Overage billed via purchased Google AI credits at Gemini Enterprise Agent Platform consumption pricing (not per-token fixed).

Featherless AI provides serverless access to LLMs and agent runtimes. It offers subscription-based 'Chat' plans with unlimited tokens and credit-based 'Developer' plans for API usage. Details were checked against the provider's own page at featherless.ai.

Free tier available — no credit card required

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.

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.

OpenAI Codex CLI — OpenAI's terminal coding agent (also web/IDE). Authenticates via ChatGPT OAuth or an API key. Model lineup (as of mid-2026): GPT-6 Astra, GPT-5.6 (Sol/Terra/Luna), GPT-5.5, GPT-5.4 / 5.4 mini, plus gpt-oss-120b; GPT-5.3-Codex-Spark is a Pro-only research preview. Plans: Free, Go ($8), Plus ($20), Pro 5x ($100) / 20x ($200); API-key mode is pay-as-you-go at API rates.

ChatGPT plans use credit-based usage allowances (credits per 1M tokens, e.g. GPT-5.6 Sol 100/10/500); not a fixed $ credit value. API-key mode billed at standard OpenAI API token rates.

Regolo AI provides scalable serverless AI infrastructure with a focus on privacy and zero data retention. It serves popular open-source models (Llama, Mistral, Qwen) via an OpenAI-compatible API, offering monthly token capacity plans and a flexible free trial.

No public free-tier information found in the source reference; treat free-tier availability as unverified rather than confirmed absent.

Speka is a unified AI API gateway providing access to frontier models (DeepSeek, Nemotron, Llama) via a single API key. It uses a subscription model with monthly usage allowances and standard per-token overage rates.

No public free-tier information found in the source reference; treat free-tier availability as unverified rather than confirmed absent.

Weights & Biases Serverless Inference gives API and playground access to open-source LLMs hosted on CoreWeave, using an existing W&B account and no separate provider keys, with support for bringing your own LoRA weights for fine-tuned models. Token rates are listed per 1M tokens on the W&B pricing page: Meta Llama 3.1 8B at $0.22 input / $0.22 output, Meta Llama 3.3 70B at $0.71 / $0.71, OpenAI GPT OSS 120B at $0.03 / $0.17, NVIDIA Nemotron 3.5 Lightning at $0.07 / $0.20 with a $0.04 cache-hit rate, and DeepSeek V4-Pro at $1.15 / $2.55 with a $0.20 cache-hit rate. Inference spend is separate from platform subscriptions and can be capped with a monthly W&B Inference budget.

Weights & Biases states only that Inference includes 'Free credits for a limited time' on the Free plan and a $5/mo credit on Pro, with no dollar amount or expiry published for the free grant (checked 2026-09-16).