Skip to main content
Updated 2026-10-06

AI Provider Capacity: What Runs Free Today & How to Reach 100%

GigTry runs at full user-facing capacity on a zero-cost keyless tier: text generation survives two independent free providers (with automatic failover), images are unlimited keyless, and search-engine submission is fully automated. Adding any free-tier API key — Groq, Gemini, Cerebras — inserts reserved capacity in front of the keyless fallback. Check live readiness any time at /api/providers/capacity.

The generation chain is built so that no single provider outage can block a user. Requests start at the fastest reserved provider and fall back automatically through free keyless strategies. This page explains what is live right now, what each upgrade adds, and how to add it — without touching code.

How the zero-cost chain works

  1. Keyless text tier (always active)

    Production runs on two independent free strategies: the Pollinations GET endpoint and the OpenAI-compatible POST endpoint (openai-fast). If one is rate-limited or down, the other takes over automatically — text generation keeps working with zero keys and zero cost.

  2. Free-tier key tier

    Any free-tier API key you add (Groq, Gemini, Cerebras, Together, OpenRouter, Cohere) slots in between the keyless tier and paid providers. Keys added in the admin AI-keys console or as environment variables are picked up automatically on the next request.

  3. Paid tier (last resort)

    OpenAI GPT-4o-mini and Claude Haiku sit at the end of the chain. They are only reached if every free option fails, and they can be left unconfigured entirely.

Check live capacity in one tap

Open /api/providers/capacity — it probes the keyless text and image endpoints in real time, shows which optional keys are present, lists the exact env vars that raise capacity, and links each provider's free-tier signup page. No secrets are ever returned, only readiness.

CapabilityRuns keyless todayWhat an optional key adds
Text generationYes — dual keyless strategies with auto-failoverReserved capacity, higher rate limits, faster latency (Groq is the highest-impact free upgrade)
Image generationYes — unlimited anonymous generation with seed rotationLicensed stock-photo search (Pixabay/Pexels/Unsplash, all free)
Voiceover / audioServer TTS with browser Web Speech fallbackReserved server-side TTS via the same ZAI key
Video generationBest-effort (heaviest capability)Reserved video capacity via the same ZAI key
Search submissionYes — Google ping, robots.txt directive, IndexNow (Bing/Yandex/Seznam), all automated dailySearch Console API reads (service account) and Bing URL-submission quota (free key)
Env varProviderFree tierSignup
GROQ_API_KEYGroqFree tier, llama-3.3-70b, fastest inferenceconsole.groq.com/keys
GOOGLE_GEMINI_API_KEYGoogle AI StudioFree tier, gemini-2.0-flashaistudio.google.com/app/apikey
CEREBRAS_API_KEYCerebrasFree tier, llama-3.3-70bcloud.cerebras.ai
OPENROUTER_API_KEYOpenRouterFree models poolopenrouter.ai/keys
TOGETHER_API_KEYTogether AIFree starter creditsapi.together.xyz
COHERE_API_KEYCohereFree trial keysdashboard.cohere.com/api-keys
GOOGLE_SERVICE_ACCOUNT_EMAIL + GOOGLE_PRIVATE_KEYSearch Console APIFree service accountsearch.google.com/search-console
BING_WEBMASTER_API_KEYBing Webmaster APIFree with Microsoft accountbing.com/webmasters
GOOGLE_PAGESPEED_API_KEYPageSpeed Insights API25k queries/day freeconsole.cloud.google.com
PIXABAY_API_KEY / PEXELS_API_KEY / UNSPLASH_ACCESS_KEYStock mediaAll freepixabay.com/api/docs/

Adding a key (two ways, both zero-downtime)

  1. Environment variables

    Add the variable in Vercel → Settings → Environment Variables, then redeploy or wait for the next deployment. The provider chain reads env on cold start and picks the key up automatically.

  2. Admin AI-keys console

    Operators can paste keys in the admin AI-keys console — they are stored server-side and loaded by the generation chain on the next request, no redeploy needed.

  3. Verify

    Open /api/providers/capacity after adding: the key shows under keysPresent and the summary reflects the new reserved provider. No code changes are ever required.

The site never depends on an upgrade

Upgrades add headroom — reserved capacity, higher limits, richer analytics. Every user-facing flow works end-to-end with zero keys: generation, images, posting pipeline, SEO automation. That is what makes the zero-cost architecture safe to run.

Frequently asked questions

Do I need to add any API keys to use GigTry?

No. Text and image generation run on a keyless free tier with automatic failover between independent providers, and search-engine submission is fully automated. Keys only add reserved capacity and higher rate limits — they are optional upgrades, not requirements.

What single upgrade gives the most capacity?

A free Groq key (GROQ_API_KEY). Groq runs llama-3.3-70b on dedicated hardware with generous free-tier limits and the fastest inference of any provider in the chain, so it absorbs traffic bursts that would otherwise use the keyless tier.

How do I check what is live right now?

Open /api/providers/capacity. It shows real-time probe results for the keyless text and image endpoints, the active fallback chain, which optional keys are present, and the exact signup links for each upgrade.

What happens if a free provider goes down?

The chain fails over automatically: the GET keyless endpoint falls back to the POST keyless endpoint, then to any configured free-tier keys, and finally to paid providers if configured. A single provider outage is invisible to users — generation retries happen server-side within the same request.

Is anything sent to paid APIs without my knowledge?

No. Paid providers are only configured when you explicitly add their keys, and they sit last in the chain. With no paid keys configured (the default), every request stays on free infrastructure.

Was this guide what you needed?