NanoGPT Review: Pricing & Comparison

A crypto-first multi-model aggregator

NanoGPT (nano-gpt.com) sits in OpenRouter’s category: a chat UI plus image/video/audio generation plus an OpenAI-compatible API, backed by prepaid balance with an optional subscription. The site says pay-as-you-go with no subscription required, and claims one OpenAI-compatible API covering 1,189 text, image, video and audio models — the widest span of any aggregator covered on this page.

Prices are published per model, with some open models well below official rates:

Selected pricing (per 1M tokens, input/cache-read/output, USD):
- DeepSeek V4 Flash 0731: $0.05 / $0.01 / $0.16 (1M context)
- GLM 5.3 Flash: $0.10 / $0.03 / $0.30
- DeepSeek V4 Flash: $0.14 / $0.0028 / $0.28
- Kimi K2.6: $0.50 / $0.13 / $2.60
- GLM 5.3: $0.60 / $0.20 / $1.10
- DeepSeek V4 Pro: $1.10 / $0.11 / $2.20
- Gemini 4 Argon: $2.00 / $0.10 / $10.00

Minimum top-ups: $0.10 in crypto or $1 by card, and crypto payments earn a 3% balance bonus across 29 supported cryptocurrencies (including Monero, Bitcoin, Nano, USDC and USDT). The FAQ states plainly that no KYC or ID check is required and that you can use the service without registering an email — which is the biggest structural difference from a typical relay.

Referrals: detailed, including the exclusions

Its invitation programme is the most specific of the vendors added in this round: friends get 5% off, and the referrer earns 10% of the invitee’s usage charges plus 5% of paid subscriptions, credited daily and payable in USD or Nano. What matters more is that it also states what earns nothing: “requests billed with provider-selection pricing or markup get no referral discount or commission. This includes Caching mode (caching: true or a :caching suffix) and pinned providers,” while automatic routing and automatic provider caching stay eligible.

That exclusion is material — if you or your invitees habitually pin providers or enable caching, real earnings will be well below the advertised 10%.

Risks you take on yourself

No SLA, no availability number. There is no uptime or latency commitment anywhere on the site. As a crypto-first resale aggregator, model provenance, upstream quota and rate-limit policy are all opaque, which means there is no guarantee that a given model will not slow down or disappear tomorrow.

The catalogue includes experimental entries, and models get retired. The site explicitly flags items such as “GLM 5.3 Uncensored — may show unusual behavior or inconsistent response quality”, and the changelog logs retirements (Sora 2 Retired on 2026-10-14). The price table also lists a “Gemini 4 Argon” entry; as of our own September 2026 verification of Google’s Gemini line-up, no Gemini 4 entry exists in the official catalogue, so third-party naming like this should be checked before integrating.

Great as a testing/fallback channel, risky as a primary. No KYC, a very low entry barrier, enormous coverage and pay-as-you-go pricing make it well suited to comparative testing and multi-model experiments. Pinning production traffic to a specific model ID, however, means accepting that there is no SLA.

Information verified 2026-10-15. Pricing, payment methods, referral terms and the model list reflect what nano-gpt.com, its API docs and its FAQ displayed on that date.

Related reviews

• OpenRouter: a cross-vendor multi-model aggregation platform, 200-300+ models, one API key to call OpenAI/Anthropic/Google/Meta/Mistral and more
• MegaLLM: 70+ models aggregated through official channels, stable and highly available, a global AI API relay for mid-to-high-end needs
• UnoRouter: specialized for role-play use cases, 200+ models, low-latency routing, 0% markup
• Requesty: EU-friendly, 400+ models, GDPR compliant, 20ms failover