GPTunneL

Positioning: a third-party model aggregation layer that doubles as a "super app"

GPTunneL is a multi-model access platform for the Russian-speaking market, and its site headline brands it as "ИИ-суперапп". Its core business is third-party model aggregation and a relay gateway: under the license offer (Оферта), models come from third-party developers called "Поставщик", and the platform only publishes the catalog, routes traffic and bills for usage, disclaiming generated content as is and taking no responsibility for the availability of supplier services. It does claim an in-house Grom line and its own inference cluster, but that is vendor self-description and currently has no independent third-party verification. It is more accurate to read it as "a multi-model gateway with an unusually heavy product layer" than as "a foundation-model company".

Dual-protocol access: two compatibility paths

• OpenAI-compatible: /v1/chat/completions supports streaming, message history and function calling, so swapping the base_url and key lets you reuse your existing SDK.
• Anthropic-compatible: /v1/messages is a genuine proxy — request and response bodies follow the Anthropic documentation, so the official SDKs and Claude Code only need a new address and key; the endpoint accepts Claude models only and returns 404 for anything else.
• Supporting endpoints: embeddings, RAG, text moderation, speech transcription, balance lookup, media generation (async jobs plus webhooks), Assistants and team quota management.

Three product lines: one account, one balance

1. Super app for creators and teams: third-party models cover text, image, video and music, alongside tools for background removal, face swap, upscaling, watermark removal, voice-over and audio/video transcription, plus Workflow canvas orchestration, one-sentence page generation with CodeBlocks, Agents and a website AI support agent.
2. API platform for developers: a separate entry point and its own documentation site.
3. Enterprise "Нейро-офис": seats at no extra charge, a shared balance, quotas by employee and department, SSO and Active Directory, automated УПД invoices via ЭДО, 223-ФЗ tender paperwork, and on-premise deployment.

Pricing and billing: prepaid balance, no monthly fee

No subscription and no mandatory payment — you top up first, then spend. The personal minimum top-up is 50 ₽, a business account is 3000 ₽, and the international edition has its own price list starting at $5. Text is billed per 1000 tokens, with requests and responses counted separately; images per image, video per second, music per track, voice-over per 1000 characters and transcription per minute. Grom 1.5 (8K context) is labelled "Бесплатно" and works without a top-up, making it the only model on the site that genuinely runs for free. There is also a tiered loyalty discount based on the last 30 days of spend, ranging from 5%–25%. The ₽ and $ price lists are fully independent, and the site offers no currency conversion of any kind.

Who it is for

Individual creators and small-to-mid teams in the Russian-speaking market who can only pay with a ruble card and still want mainstream models plus image, video and music from one balance; Russian companies or foreign entities operating in Russia that need contracts, УПД, per-person quotas and audit of employee usage; and developers who already have OpenAI or Anthropic code and want to move to a ruble billing channel with almost no rework. Conversely, it is a poor fit if you need something permanently free, a prepaid plan that never expires, or large single-batch file processing, or if you are highly sensitive to liability caps and data jurisdiction.

Listing notes and risk disclosure

The hard constraints in the terms are worth reading first: the platform may suspend service when an upstream supplier blocks requests; it may unilaterally ban a customer without explaining why; compensation is capped at the amounts already collected in the previous billing period; disputes are governed by Russian law and heard by the court at its registered address; and discount Bonus expire 30 days after crediting, refundable only when "completely unused". On availability, in 2026-10 it temporarily cut off all traffic from outside Russia after a large-scale DDoS attack and opened the service to Russian IPs only, while payment channels also dropped out briefly in the same period — mission-critical workloads should carry their own redundancy. On price, a third party estimates that at GPT-4o rates its unit prices run about 3–4× the official API. This listing covers the aggregation layer itself and does not mean its upstream relationships are licensed by the model vendors; prices and the model catalog change in real time, so before integrating, rely on the personal pricing you see after logging in and on the public documentation.