Settings
Providers, keys, model selection, preferences, data. All local.
ai providerswhich one should I use? →
0 configuredFree tier serving Llama / Mixtral at the fastest tokens-per-second on the market.
Free tier: ~30 requests/min, ~6,000 tokens/min. Plenty for everyday use.
▸ rate limits · free
- ·Free tier: 30 RPM, ~14,400 RPD, 6,000 TPM on the larger models.
- ·The 20B models have higher TPM (30,000) for short tasks.
- ·Hard quota — exhaust the daily cap and you wait until midnight UTC.
- official docs
image upload disabled on this provider — switch to Anthropic, OpenAI 4.1+, Gemini, or OpenRouter
Fastest tokens-per-second on the market, on specialized hardware.
Free tier available; rate-limited but usable for everyday generation.
▸ rate limits · free
- ·Free tier: 30 RPM, 60,000 TPM.
- ·Fastest tokens-per-second on the market (specialized hardware).
- ·Daily-token limits not publicly published — refresh on rate-limit error.
- official docs
image upload disabled on this provider — switch to Anthropic, OpenAI 4.1+, Gemini, or OpenRouter
One key → access to many models including FREE community models. Best for trying multiple models without multiple keys.
Models tagged ':free' have generous free quotas — the fastest way to run OpenAdKit at $0.
▸ rate limits · free
- ·Models tagged ':free' (Gemma 4, Nemotron 3): 20 RPM, 50 RPD.
- ·Free-tier accounts may face additional caps based on credit balance.
- ·Paid usage routes via your OpenRouter credit — buy credits separately.
- official docs
Strong at multimodal + long-context. Generous free tier across the Gemini 3 family.
Free tier (AI Studio key): rate-limited but genuinely usable on the Flash + Flash-Lite models.
▸ rate limits · free
- ·Free daily caps are PER MODEL and vary a lot — gemini-3.6-flash is ~20 req/day.
- ·Flash-Lite models have much higher free caps — prefer them for volume.
- ·429 'quota exceeded' = day or minute cap. Look at the retry-in seconds.
- ·Paid tier (with billing): 2000 RPM, no daily cap, 4M TPM.
- official docs
Hosts hundreds of open-source models. Pay-per-use; some free models available.
Some models have free quotas via the dashboard. Most are pay-per-use at cheap rates.
▸ rate limits · paid
- ·Most models pay-per-token. Cheap rates (~$0.15/M for GPT-OSS 120B).
- ·A handful of 'free' models have daily quotas — check the model card.
- ·$5 free signup credit; expires.
- official docs
image upload disabled on this provider — switch to Anthropic, OpenAI 4.1+, Gemini, or OpenRouter
The model OpenAdKit was originally built for. Strongest at long-context reasoning + safety. No free tier; pay-per-use.
▸ rate limits · paid
- ·Pay-as-you-go, no monthly minimum after first $5 top-up.
- ·Default tier-1 limits: 50 req/min, 40k input tokens/min for Sonnet.
- ·Higher tiers unlock automatically based on spend history.
- official docs
GPT family. Strongest at structured-output JSON. No free tier — pay-per-use.
▸ rate limits · paid
- ·Tier-1 (≥$5 paid, <7 days old): 500 req/min for GPT-4.1 / GPT-5.
- ·Limits scale with billing history — tier-5 is 10,000 req/min.
- ·GPT-4.1-mini: cheapest paid tier, ~$0.40/M input.
- official docs
Cheapest serious reasoning model out there. Pay-per-use, no free tier.
▸ rate limits · paid
- ·DeepSeek V4 Flash: ~$0.14/M input, ~$0.28/M output. No free tier.
- ·DeepSeek V4 Pro: ~$0.44/M input, ~$0.87/M output.
- ·Default RPM: tier-based, ~60 RPM for new accounts.
- official docs
image upload disabled on this provider — switch to Anthropic, OpenAI 4.1+, Gemini, or OpenRouter
European AI lab. Strong at multilingual + JSON output. No free tier, but inexpensive.
▸ rate limits · paid
- ·Mistral Large: ~$2/M input, $6/M output.
- ·Mistral Small: ~$0.20/M input, $0.60/M output — recommended cost/quality.
- ·Default 60 RPM; multilingual + JSON output is a strength.
- official docs
image upload disabled on this provider — switch to Anthropic, OpenAI 4.1+, Gemini, or OpenRouter
preferences
Applied to budget inputs in optimizers + launch wizard, and to AI cost previews. Exchange rates are coarse approximations — accurate enough for previews, not accounting.
OpenAdKit uses Jina Reader to pull a website's content when you add a brand by URL. The free tier hits ~20 req/min limits and may 401 on bursts; paid keys lift that. Get one at jina.ai/reader. AllOrigins is used as a free fallback automatically.
monthly budget
Calculated from your saved history, so it reflects real generation cost — not the lifetime counters below. You'll see a warning at 80%.
usage
counters live in this browser only. real billing on your provider's console.
data
backups include brand brains + history. api keys NOT included.
key encryption
Off by default: keys live in plain localStorage, readable by any browser extension or anyone with access to this profile. Turning this on encrypts them with a passphrase you enter once per session. There is no recovery — forget the passphrase and you re-enter your keys.
danger zone
wipe everything: keys, brains, history, usage counters.