Most of your LLM spend is mis-routed to models that are overkill.
Most teams send every query to Sonnet or GPT-4o. A third of those are formatting, syntax questions, regex explanations, or boilerplate generation - tasks an 8B or 70B open-weight model handles just as well at <5% of the cost.
✗ Today (single-model)
Every request hits Claude Sonnet 4.6 at $3 input / $15 output per 1M tokens. You pay frontier rates for the long tail of work that doesn't need it.
✓ With llmdeal.me
Each request goes to the smallest open model that can handle it. Simple stuff runs on an 8B model for cents, hard stuff gets a frontier open model, and you only pay frontier rates when the work actually needs it.
Why it costs less.
No magic, no VC hand-waving. Just a few boring infrastructure choices that add up.
01Smart routing
A classifier scores each request and sends it to the cheapest model that can do the job. You can tune how aggressive it is, or just pin a specific model when you want one.
02Open models, hosted by us
We run open-weight models on our own GPU. Live right now and free on every plan: Dolphin 2.6, NeuralHermes 2.5 and Qwen 3.6 35B. Bigger models like Llama 3.3 70B, Qwen3-235B or GPT-OSS-120B route on request. Closed models such as Claude or GPT run through your own key or credits, never resold with margin.
03US & EU capacity
GPU capacity in both the US and the EU, open to developers anywhere. Need inference to stay in the EEA? Flip on EU-resident routing and it does.
04Crypto and privacy
Pay in BTC, XMR or LTC, no KYC. We keep only what billing needs and never store your prompts, responses or IPs. GDPR rules for everyone, everywhere.
The three core tiers
Subscribe monthly for the included token allowance, or buy pay-as-you-go credits at the per-1M rate. Same key, same models. Full pricing on the pricing page.
Starter
or PAYG · $0.60 / $1.20 per 1M tokens
- Llama 3.1 8B on Groq + cheap smart-routing
- OpenAI-compatible API · 32k context
- EU-resident routing (opt-in)
- Overage: $5 per extra 1M tokens
Pro
or PAYG · $2.00 / $5.00 per 1M tokens
- Full open-model mix · 128k context
- Per-request cost telemetry
- Priority routing during peak load
- Overage: $4 per extra 1M tokens
- Everything in Starter
Elite
or PAYG · $4.00 / $9.00 per 1M tokens
- Top open models · 128k context
- EU-resident routing on by default
- Dedicated rate-limit pool
- Direct DM support (Matrix / Telegram)
- Everything in Pro
Smaller budget? Plans start at $9/mo, and a $5 trial pack gets you ~5M tokens that never expire. Bigger? Business and Scale go up to 1.5B tokens/mo. Every plan runs the open-model pool; closed models (Claude, GPT, Gemini) are available via your own key or credits. It's all on the pricing page.
Try llmdeal today.
Pick a subscription tier or buy PAYG credits, pay in BTC / XMR / LTC, get an API key back. No card, no KYC, no commitment past the current month. Not ready to commit? The $5 trial pack gets you ~5M tokens that never expire.
Talk to the operator directly.
Three channels. Matrix is preferred (E2EE, federated, no phone number required). Telegram works. Discord works too, but the invite rotates - DM for the current link if the public one has expired.
- Matrix (preferred): handle rotates and isn't printed hereDM on Discord or email and you get the current one back.
- Telegram: same deal, handle on requestDM on Discord or email for the current one.
- Discord (rotating invite): discord.gg/ZgcKssAWJ6If the link is dead, email for a fresh one.
- Email: [email protected] · [email protected]
Privacy & data
We run under GDPR (Norway / EEA) and apply the same rules to everyone, wherever you are. No legal boilerplate, just the facts.
✓ What we store
Order record. Order ID, SKU, currency, amount, status, timestamps. Append-only ledger - needed to credit your account and reconcile payments.
Contact handle. The email or messaging handle you provide at checkout. Used only to deliver your API key and billing notifications.
Token counts. Once live, we log input + output token counts per request for billing. The prompt content is never stored.
✗ What we don't store
Prompt content. Prompts and responses are discarded when the request completes. No training. No content audit log.
KYC / identity data. No name, address, government ID or card details, ever. Crypto only.
IP addresses. Held transiently for fraud and rate-limit checks, deleted within 24 h of order settlement.
Want your data deleted? DM us from the handle you signed up with and we remove the order record within 48 hours, timestamp included. That's GDPR Article 17 for every customer, EU or not. Full policy: /privacy.html.
FAQ
Straight answers. DM if anything's unclear.
How is this 55-85% cheaper than my current provider?
Most apps send everything to Sonnet or GPT-4o at frontier prices, but a big share of those requests are simple - formatting, syntax, short answers - and a small open model handles them just as well for a fraction of the cost. Smart routing sends each request to the smallest model that can do the job.
Check the per-token prices on Artificial Analysis and quality on LMSYS Arena before you buy.
Subscription or pay-as-you-go - which should I pick?
Predictable monthly usage? Subscribe - plans run from $9/mo up to 1.5B tokens/mo, cancel any time, no card stored. Spiky usage or just trying it out? Buy PAYG credits from $20; they sit on your account until you spend them.
Same key works for both, so you can stack a subscription with a PAYG balance for overage. Full list on the pricing page.
What models can I use right now?
Hosted on our own GPU, free on every plan: Dolphin 2.6-Mistral, NeuralHermes 2.5-Mistral and Qwen 3.6 35B. These open-weight models are live right now (a few are uncensored variants, fitting the no-KYC ethos).
On request: bigger open models like Llama 3.3 70B, Qwen3-235B and GPT-OSS-120B route through smart-route or by passing the name directly - ask and we'll switch them on. Full catalog on the models page.
Can I use Claude, GPT or Gemini?
Yes, but we don't host those - they're closed models. Two ways to reach them: bring your own key (BYO - we charge $0 for the proxy hop, you pay the provider directly) or pre-purchase credits on a subscription.
Everything we host ourselves is open-weight, so there's no closed-model reseller margin riding on your tokens.
How do I pay?
BTC is automatic - go to /buy.html, pick a plan, get an address, send, and your key is DMed back on confirmation. XMR and LTC are semi-manual: pick the plan, DM the operator for a one-time address, and your account is credited within a few hours.
No KYC on any of them. Cards need KYC and invite chargebacks, and devs paying for an API shouldn't have to identify themselves to use one.
What data do you keep about my prompts?
Just the order record, your contact handle and token counts. Prompt and response content are never stored. Full breakdown in the Privacy & data section above.
Requests routed to upstream providers fall under their own policies, same as calling them directly - and none of them train on API traffic by their stated policies.
Who runs llmdeal.me?
One EEA-based independent operator. No VC, no team, no roadmap deck - just direct DM support on Matrix, Telegram or Discord. See the contact section.
