OpenRouter alternative for crypto and no-KYC open-model inference

WeightsAPI serves 15 open-weight models behind one OpenAI-compatible endpoint, https://weightsapi.com/v1: Llama, Qwen3, DeepSeek, Gemma 3, Mistral and Mixtral, plus the fine-tunes Hermes 4 70B and Dolphin Mistral 24B Venice Edition. Each model has its own input and output price per million tokens, published on the pricing page and in prices.json.

The main benefit over an aggregator is the account and payment model. One balance, funded in the coin you already hold, pays for per-token calls and for monthly dedicated GPUs from $416 (L40S) to $9,736 (8 × H100), with no card, email or identity document in the current signup and payment flow.

1. Pay in Monero or seven other coins, with no platform fee

Top-ups accept XMR, BTC, ETH, TRX, USDT on TRON or Ethereum, USDC on Ethereum, SOL and LTC, and we add no platform deposit fee: you pay only the network fee on your transfer, and the credit lands within about a minute once the network confirmations are reached. OpenRouter accepts crypto in USDC only and charges 5% on crypto purchases, about $5 per $100. See every asset and confirmation count.

2. A username instead of an email address

Signup asks for a username and a password of at least 15 characters: no email, phone, ID or card. OpenRouter’s terms ask for an email address or other contact information. A one-time recovery code replaces the reset email. Read what the no-KYC account covers.

3. Lower prices on Mixtral 8×22B and Llama 3.1 8B, cheaper output on three more

We beat OpenRouter’s listed prices on Mixtral 8×22B ($0.60 / $0.60 per million tokens against $2.00 / $6.00), Llama 3.1 8B ($0.03 / $0.05 against $0.05 / $0.08), and on output for Gemma 3 27B ($0.15 against $0.45), DeepSeek V3 ($0.85 against $1.00) and Qwen3 32B ($0.20 against $0.28). Gemma 3 27B comes out cheaper once output passes about 6% of your tokens, DeepSeek V3 at any mix, Qwen3 32B only above about 47%. See the full table, losses included.

4. A monthly GPU on the same balance

The same crypto balance prepays a dedicated GPU for one month from activation: an H100 SXM 80 GB costs $1,352 a month, with no per-token charges on that machine. OpenRouter has none; RunPod lists an H100 SXM pod at $3.49 an hour, about $2,548 over 730 hours, while ours is managed inference capacity without shell access by default. Compare the six GPU configurations.

5. Hermes 4 70B, and no filter added at the gateway

Hermes 4 70B ($0.13 / $0.40) is in our catalog; on 25 September 2026 it was absent from OpenRouter’s public model list, which carries Hermes 4 405B instead. We add no content filter at the gateway, so fine-tunes such as Dolphin Mistral 24B Venice Edition answer as trained; their own training can still refuse. See the models for creative and roleplay work.

6. A spending cap on every key

Each API key can carry a lifetime USD spending cap and a model allowlist, and can be revoked at any time, so every app gets its own limit. See how keys and caps work.

Competitor prices come from each provider’s public model list on 25 September 2026 and can move; ours come from prices.json (updated 24 September 2026).

We are not the cheapest everywhere. Llama 3.3 70B, Qwen3 235B A22B, Mistral Small 3 and DeepSeek V4 Flash ($0.049 / $0.098 against our $0.09 / $0.18) cost less on OpenRouter, and NanoGPT leads on DeepSeek V3 and R1.

At an 80/20 mix of 16 million input and 4 million output tokens, Gemma 3 27B costs $2.20 here, $3.08 on OpenRouter ($3.23 with its 5% crypto fee) and $5.98 on NanoGPT. At the same mix Qwen3 32B is cheaper on OpenRouter ($2.40) and NanoGPT ($2.80) than here ($3.20). On Mixtral 8×22B, 10 million input plus 2 million output tokens cost $7.20 here against $32.00 on either aggregator (our calculations). Browse all 15 models and their context limits.

Also weighing NanoGPT? Here is what crypto buyers check first, from each provider’s own pages.

In short: NanoGPT is the lightest entry point and OpenRouter the widest catalog. WeightsAPI is the pick when you want Monero or Bitcoin funding with no platform fee, an account without contact data, the lowest price of the three on Mixtral 8×22B and Llama 3.1 8B, Hermes 4 70B, and a dedicated GPU on the same balance.

Full rules: billing and crypto deposits, and the terms.

If your code reaches OpenRouter through the OpenAI SDK, moving a workload takes three edits: base URL, API key and model ID. Our model IDs keep the publisher’s spelling, such as Qwen/Qwen3-32B, so copy them from the models page.

-client = OpenAI(base_url="https://openrouter.ai/api/v1", api_key=os.environ["OPENROUTER_API_KEY"])
+client = OpenAI(base_url="https://weightsapi.com/v1", api_key=os.environ["WEIGHTSAPI_API_KEY"])
-model = "mistralai/mixtral-8x22b-instruct"
+model = "mistralai/Mixtral-8x22B-Instruct-v0.1"

Chat Completions, Completions and the model list follow the OpenAI format; strip OpenRouter-only options such as provider routing. Once your model shows as serving on /status, test a sample and keep your OpenRouter key as a rollback path. See the migration guide and the SillyTavern setup guide.

Why USD 20 when OpenRouter takes $5? Each top-up is its own on-chain order, and a larger transfer keeps the network fee a small share. It is a per-top-up minimum, not a balance floor, and with no platform fee a $100 top-up becomes $100 of credit, where $100 of OpenRouter credit bought in USDC carries about $5 in fees.

What will a month cost? Your tokens at the published rate plus the network fee on each top-up: 20 million input and 5 million output tokens on Gemma 3 27B come to $2.75. No subscription, no platform fee, and a lifetime cap on each key sets the ceiling.

Is my model serving? Check it on the status page before topping up; dedicated GPUs get their activation date confirmed before you pay.

What does no KYC cover? No ID, email or phone in the current account flow. It is not anonymity: keys, usage and payment records stay linked to your account, and late or different-amount transfers go to review.

How long until credit shows? Within about a minute once the network confirmations are reached (1 on Bitcoin, 10 on Monero, 19 on TRON), so you can start as soon as your transfer confirms.

Refunds and trust? Credit cannot be withdrawn to a wallet; a GPU order cancelled before activation is refunded to your balance. OpenRouter never refunds crypto either. Rates, billing rules and the list of records we keep are published, so start with one top-up and a capped key.

Choose WeightsAPI if you pay in Monero, Bitcoin or another coin and want no platform fee; you prefer an account without email; you run Mixtral 8×22B, Llama 3.1 8B, Hermes 4 70B, Gemma 3 27B or output-heavy Qwen3 32B; or you plan a monthly GPU.

Stay with OpenRouter, or pick NanoGPT, if you need GPT, Claude or hundreds of models behind one key; you top up under $20 at a time; you mainly run Llama 3.3 70B, Qwen3 235B A22B, DeepSeek V3 or R1, or input-heavy Qwen3 32B; or you need a contractual SLA, which WeightsAPI does not offer.

What to do now: create a username account and store the recovery code offline, check your model on the status page, top up from USD 20, and point your base URL at https://weightsapi.com/v1. Planning a GPU? Pick a configuration and get the activation date confirmed before you pay.