Free first · then paid · every claim sourced

Best AI Model API Providers

Gateways, routers, and direct model APIs ranked free tier first, then paid. Free access types are labeled honestly — always-free, monthly credit, signup credit, or trial. Every row links an official source and a real verification date. We do not invent free tiers.

20 free-tier providers9 paid providersVerified 2026-09-23

Free tier — ranked strongest free access first

Google Gemini API

Direct APIAlways free

Free input & output tokens on multiple Gemini models via AI Studio

Free

Free tier: free input and output tokens on multiple models, free context caching, free AI Studio access. Rate-limited; some data used to improve Google products.

Paid pricing

Paid tiers after free allotment; Flash-class intro pricing $0.75 / $3.75 per 1M input/output through Dec 31, 2026 (ai.google.dev).

Gemini family (limited model set on free tier)frontier free usageprototypinglong context

Caveat: Free tier rate-limited; direct page fetch intermittent — verify limits in AI Studio console.

Official siteSource pageVerified Sep 23, 2026LIVE

Cloudflare Workers AI

Cloud inferenceAlways free

10,000 Neurons/day free on every Workers plan — no credit card

Free

10,000 Neurons per day free (resets 00:00 UTC) on Free and Paid Workers plans. Beyond free: $0.011 / 1,000 Neurons. Some frontier models require a paid billing method.

Paid pricing

Workers Paid plan; usage $0.011 / 1,000 Neurons above free allocation.

LLM, image, audio, embedding models (per-model neuron rates published)edge inferencedaily free allotmentworkers stack
Official siteSource pageVerified Sep 23, 2026LIVE

OpenRouter

Gateway / routerAlways free

50 free requests/day on 25+ free models — one API, 500+ models

Free

Free plan: 25+ free models, 4 free providers, chat + API, 50 requests/day, no payment options required. `:free` model IDs rate-limited; Free Models Router (`openrouter/free`) available.

Paid pricing

Standard pay-as-you-go: 5.5% platform fee on provider pricing (Business 8%); credit card, crypto, and more. 500+ models, 80+ providers.

500+ models, 80+ providers (paid); 25+ free models on free planmulti-model gatewaytrying free modelsone API key
Official siteSource pageVerified Sep 23, 2026LIVE

OrcaRouter

Gateway / routerAlways free

Free forever Hacker tier: 200+ models, 0% token markup, no card

Free

Hacker plan free forever: full gateway (200+ models, OpenAI-compatible API, auto-failover), 10 API keys, 0% token markup, no credit card.

Paid pricing

Team and Enterprise custom (feature subscriptions); tokens always billed at upstream provider list rate with no OrcaRouter markup.

200+ models across OpenAI, Google, Qwen, Anthropic, DeepSeek, and otherszero-markup routingfree gatewayfailover
Official siteSource pageVerified Sep 23, 2026LIVE

Requesty

Gateway / routerAlways free

200 free requests/day on free models — 600+ models, 5% PAYG markup

Free

Free plan $0: all free models, 200 requests/day, routing/caching/fallbacks, analytics, EU data residency, no credit card.

Paid pricing

Pay-as-you-go: model cost + 5% markup (bring-your-own-keys). Enterprise custom.

600+ models, 20+ providersproduction gatewayEU residencyfree daily quota
Official siteSource pageVerified Sep 23, 2026LIVE

TokenHarbor

Gateway / routerAlways free

Free $0 plan with monthly free-model allowance — no card required

Free

Free plan $0/month: free monthly allowance (4-week cycle), rotating free models (e.g. DeepSeek V4.1 Flash, MiMo V2.6 Flash, TH-Rudder), one OpenAI-compatible API, no card. Free models must be enabled in dashboard; free routes never charge balance.

Paid pricing

Pay-as-you-go wallet (top-ups from $10) + optional passes: Agent $0.99 first month then $1.99/mo ($10 usage); Office $9.99/mo ($35); Frontier $99/mo ($180). 0% token markup currently.

9+ premium models + free set (official marketing; exact catalog size not hard-counted)free allowanceagent APIone wallet

Caveat: Some marketing pages mention welcome credit; official docs say accounts start at $0 with no welcome credit.

Official siteSource pageVerified Sep 23, 2026LIVE

TokenRouter

Gateway / routerAlways free

Free gateway: 1,000 requests/month, 2 of 13 providers, BYOK, zero markup

Free

Free plan $0 flat forever: 1 seat, 1 team, 2 connected providers, 1,000 gateway requests/month, budgets/hard caps/analytics, no credit card.

Paid pricing

Starter $29/mo (1M requests, all 13 providers); Team $99/mo (10M); Enterprise $499/mo — flat subscription, zero token markup (providers bill at exact rates).

13 providers (BYOK gateway)spend controlsBYOK gatewayfree monthly quota

Caveat: Different product from tokenrouter.com (which showed an empty model catalog on 2026-09-23).

Official siteSource pageVerified Sep 23, 2026LIVE

Mistral API

Direct APIMonthly credit

Free plan includes $10/mo in API credits

Free

Free plan: $10/mo in API credits, limited Vibe/Studio access, test Mistral models in Studio. PAYG credits can extend past plan limit.

Paid pricing

Pro $14.99/mo ($30/mo API credits); Team $24.99/user/mo ($50/mo credits); Enterprise custom. Most models priced per million tokens.

Mistral family (Medium, Small, OCR, Voxtral, open-weight models)monthly free creditsEU provideropen-weight options
Official siteSource pageVerified Sep 23, 2026LIVE

LiteLLM

Open sourceAlways free

Open-source AI gateway — free forever, self-host, 100+ providers

Free

Open-source gateway $0 forever: self-host in minutes, no credit card, 140+ provider integrations, virtual keys, budgets, load balancing, guardrails. MIT-licensed core.

Paid pricing

Enterprise: support, SSO, JWT auth, audit logs, custom SLAs (talk to sales).

140+ providers / 1,800+ models via one OpenAI-compatible APIself-hosted gatewaybudgetsno vendor lock-in
Official siteSource pageVerified Sep 23, 2026LIVE

GroqCloud

Direct APIAlways free

Free $0 plan to start on fast LPU inference APIs

Free

Free plan ($0) documented for getting started with APIs; community support; downgrade path from Developer confirmed in billing FAQ. Exact free-tier RPM/TPM not published on public docs — check console.

Paid pricing

Paid tiers with higher rate limits (Developer and above; see GroqCloud billing).

Open and partner models on Groq LPU inferencespeedfree API entrylow latency

Caveat: Do not publish specific free-tier RPM/TPM without reading console — not on public rate-limits page.

Official siteSource pageVerified Sep 23, 2026LIVE

Cohere API

Direct APITrial

Free trial keys: 1,000 API calls/month, 20 req/min chat

Free

Trial (evaluation) API keys: free, limited to 1,000 API calls/month; chat models 20 req/min; embed 2,000 inputs/min trial.

Paid pricing

Production keys paid (most chat models 500 req/min); see Cohere pricing docs.

Command A / R family, embed, rerank, parseRAGenterprise chat modelsevaluation keys
Official siteSource pageVerified Sep 23, 2026LIVE

Helicone

Gateway / routerAlways free

Hobby free: 10,000 requests + 1 GB observability, gateway at 0% markup

Free

Hobby free forever: 10,000 requests, 1 GB storage, 1 seat, 1 organization, 7-day retention. Gateway features (caching, rate limits, fallbacks) on free tier.

Paid pricing

Pro $79/mo (unlimited seats); Team $799/mo (SOC-2/HIPAA); Enterprise custom.

Multi-provider observability + gateway (OpenAI, Anthropic, Azure, LiteLLM, etc.)observabilityfree logging quotaopen-source
Official siteSource pageVerified Sep 23, 2026LIVE

Portkey

Gateway / routerAlways free

Developer free: 10k logs/month — enterprise AI gateway

Free

Developer plan: 10,000 logs/month free forever (BYOK + observability).

Paid pricing

Production $49/mo; Enterprise custom (SSO, guardrails, compliance).

100+ providers via one gateway (count not stated as fixed on pricing page)enterprise gatewayobservabilitycompliance
Official siteSource pageVerified Sep 23, 2026LIVE

Vercel AI Gateway

Gateway / routerMonthly credit

Free tier with monthly free credits on a model subset — zero token markup

Free

Every Vercel team gets a free tier: monthly free credit (amount not restated on the fetched pricing body — check dashboard), free-tier model subset, lower rate limits. Free credits start on first AI Gateway request; buying credits moves you to paid tier.

Paid pricing

Zero markup on tokens at provider list price; BYOK no fee. Paid AI Gateway Credits; volume discounts available.

Multi-provider catalog (free subset only on free tier)Vercel appszero markupfree model subset

Caveat: Free credit amount not printed on the pricing page body we loaded — verify in Vercel dashboard before quoting a $ figure.

Official siteSource pageVerified Sep 23, 2026LIVE

Hugging Face Inference Providers

Gateway / routerMonthly credit

$0.10/month free credits across 200+ models via one router

Free

Free users: $0.10/month credits (subject to change); pay-as-you-go after credits at pass-through provider rates.

Paid pricing

PRO $2.00/mo; Team/Enterprise $2.00/seat; usage pass-through after free credits.

200+ models via inference providersopen modelssmall monthly creditHF ecosystem
Official siteSource pageVerified Sep 23, 2026LIVE

Novita AI

Direct APIAlways free

Some models free ($0 input/output) plus 200+ paid models

Free

Official pricing page shows free model rows: Ling 3.0 Flash Fin, Ling 3.0 Flash VL, Ling 3.0 Flash Sante — Input Free / Output Free (256K context). Rest of catalog per-token paid.

Paid pricing

Per-token PAYG across DeepSeek, Qwen, GLM, Kimi, and others; batch inference intro 50% off on supported models.

200+ models, GPU instances, agent sandboxesfree model rowscheap open modelsGPU + API
Official siteSource pageVerified Sep 23, 2026LIVE

Fireworks AI

Direct APISignup credit

$1 free signup credit for serverless inference

Free

$1 free credits at signup (serverless inference, postpaid billing). One-time credit — not a recurring free tier.

Paid pricing

Pay-as-you-go per token after credits; postpaid billing.

Open and partner models (fast inference cloud)quick trialopen modelsserverless

Caveat: One-time signup credit only.

Official siteSource pageVerified Sep 23, 2026LIVE

Together AI

Direct APISignup credit

$150 free credits via official claim form

Free

$150 free credits via official claim form (not instant — form-based). Docs historically stated no free trial; current official page offers this claim path.

Paid pricing

Pay-as-you-go after credits; minimum $5 credit purchase, fully prepaid (balance 0 suspends access).

Full library (DeepSeek, Kimi, Qwen, GLM, image/video)large signup creditopen modelsprepaid

Caveat: Credits claimed via form; older billing docs said no free trial — verify current form terms.

Official siteSource pageVerified Sep 23, 2026LIVE

Cerebras

Direct APITrial

$5 free trial credits (30 days) — card required, no permanent free tier

Free

Free trial: $5 credits expire 30 days after grant; verified payment method required (API/Playground inactive without card). Trial rate limits published (e.g. 5 RPM on trial models). Explicitly no permanently free tier.

Paid pricing

Pay-as-you-go after trial credits; per-token pricing by model.

Fast inference on Cerebras hardware (open + partner models)speedshort trialevals

Caveat: Trial requires card; not a permanent free tier.

Official siteSource pageVerified Sep 23, 2026LIVE

PiAPI

Gateway / routerSignup credit

$0.50 free signup credits for unified generation APIs

Free

New accounts: $0.50 free credits; Free plan $0/mo (limited speed, basic APIs, playground trial).

Paid pricing

Creator $15/mo ($10 bonus); Pro $60/mo ($60 bonus); Enterprise $100/mo.

Flux, Gemini, Qwen, Kling, Veo, Sora2, LLM APIsmedia APIssmall signup creditplayground

Caveat: Small one-time credit.

Official siteSource pageVerified Sep 23, 2026LIVE

How this list works: we load each provider's official pricing or docs page, record only claims printed there, and date the check. If a free tier could not be verified on an official page, the provider is listed as paid or omitted — not softened into "free".