OpenRouter vs LiteAI — pricing, markups, stability, and taxes (updated 2026-07-22)

As of 22.07.2026, the "openrouter" cluster gets ~60,600 monthly searches in Russia (Yandex.Wordstat): openrouter 38 978, openrouter ai 13 092, openrouter api 2 490, openrouter key 1 542, пополнить openrouter 1 027. Weekly volume grew +76% in two weeks. OpenRouter is often pitched as "the universal proxy for all LLMs", and LiteAI is Anthropic's proxy for Russia-side developers. Both provide API keys and speak OpenAI-compatible format, but the difference in tariffs, markups, stability, and tax burden is significant.

This article is a side-by-side comparison across four axes: tariffs, markups, stability, and taxes, plus an "OpenRouter AI explained" section and a "OpenRouter alternatives in Russia" section. No hate, no artificial advantages, just actual data from both providers' websites and their payment processors' public reports.

What is OpenRouter AI (and why is openrouter ai a 13k/mo query)

OpenRouter AI is the same as OpenRouter — the openrouter.ai brand is sometimes written as OpenRouter AI or OpenRouterAI. The service is an AI model aggregator (router): one API key → 400+ LLMs from 70+ providers (Anthropic, OpenAI, Google, Meta, Mistral, DeepSeek, Qwen, Cohere, etc.).

What it technically does: accepts requests in OpenAI-compatible format (/chat/completions), routes by the model field ("anthropic/claude-opus-4.8", "openai/gpt-5"), translates to the provider's native API, and returns unified OpenAI-format responses.

Why RU devs search for it: one key instead of five; auto-routing; Stripe-USD billing (blocked for RU cards); free-tier on some models (meta-llama/llama-3.3-70b:free etc.). The pain point: payments — that's why openrouter в россии (601/mo) and аналоги openrouter (430/mo) exist.

TL;DR: what to choose

If you care about…	Choose
Tariff — fixed price per 1M tokens, independent of FX	LiteAI (30 ₽/M flat)
Markup — no hidden fees on top of provider list-price	LiteAI (flat price) vs OpenRouter (5.5% fee)
Stability — multi-provider failover and public status page	OpenRouter (Bedrock + Vertex + Anthropic direct)
Stability — single-backend with predictable latency and zero-data-log	LiteAI (Anthropic only)
Taxes — RUB receipt with VAT/no-VAT for Russian accounting	LiteAI (YooKassa, vat_code: 1)
Taxes — corporate USD invoice with W-9/W-8 for enterprise	OpenRouter (Stripe Invoicing)
Access to 400+ models from 70+ providers via one key	OpenRouter
Auto-routing and fallback between model providers	OpenRouter
Payment in Crypto (USDT/USDC) without bank cards	Both (LiteAI via Bitbanker, OpenRouter via Coinbase Commerce)
BYOK on your own Anthropic key with minimal fee	OpenRouter (5% fee after $25k/mo)
Non-RU/CIS team with a US/EU card	OpenRouter
OpenRouter AI as a brand (popular 13 092/mo query)	Both — both run on openrouter.ai, difference is payment path and tariff
OpenRouter alternatives in Russia (430/mo) — need RUB payment	LiteAI (Claude-only) / ProxyAPI (multi-provider, RU) / Bothub (multi-provider, RU)
OpenRouter API in Russia (2 490/mo)	Works, but needs VPN + USD card or crypto; LiteAI is more convenient for Claude-only

This is not "either/or" — both can coexist. Stack-of-stack: OpenRouter as an upper-tier gateway for non-Anthropic models and auto-fallback, LiteAI as the primary Claude backend for Russia.

---

1. Tariffs

LiteAI: flat price per token

LiteAI sells tokens in packages in RUB or USD (via Bitbanker). The tariff is flat for all Anthropic models, except fable-5:

Package	RUB Price	USD Price	Tokens	Cost per 1M tokens
1M	30 ₽	$1	1,000,000	30 ₽ / $1 per 1M
5M	150 ₽	$5	5,000,000	30 ₽ / $1 per 1M
100M	2,000 ₽	$70	100,000,000	20 ₽* / $0.70 per 1M
10M (fable-5)	5,000 ₽	$100	10,000,000	500 ₽ / $10 per 1M

* The 100M package provides a discount: 2,000 ₽ for 100M = 20 ₽/M. A ~33% volume discount.

Key features:

• No input/output split. Anthropic charges 5× more for output. LiteAI charges the same for input and output.
• Model agnostic. 30 ₽ per 1M applies to both Opus 4.8 and Haiku 4.5. LiteAI "averages out" the cost, which is a bargain for Opus and neutral for Haiku.
• No expiration. The sk-bf-… key lasts until tokens are depleted.
• Fair-use rate-limit. No hard RPM limits, protected by anti-abuse + Anthropic-side 429 during spikes.

OpenRouter: provider list-price + fee

OpenRouter does not set its own tariff on models — it takes the provider's list-price and adds a platform fee:

Element	Value
Model base price	= Anthropic / OpenAI / Google / etc. official price
Platform fee (PAYG)	5.5% on inference cost
Platform fee (BYOK)	0% for first $25k/mo, 5% after
Minimum top-up	$5 (PAYG), $50 for crypto
Billing	Input and output separate, per provider rates

Example for Opus 4.8 (Anthropic) on 100M tokens mixed (80% input / 20% output):

Element	Cost
Input: 80M × $3/M	$240
Output: 20M × $15/M	$300
Subtotal	$540
Platform fee 5.5%	$29.70
Total OpenRouter	$569.70

LiteAI takes 3,000 ₽ for the same 100M (approx. $33.30 at 90 ₽/$ rate) — roughly 17× cheaper, even including the 5.5% fee.

1M token price comparison

Model	Anthropic direct (in / out)	OpenRouter (incl. fee)	LiteAI (flat)
Claude Opus 4.8	$3 / $15	$3.165 / $15.825	30 ₽
Claude Sonnet 4.6	$3 / $15	$3.165 / $15.825	30 ₽
Claude Haiku 4.5	$1 / $5	$1.055 / $5.275	30 ₽
Claude fable-5	$5 / $25	$5.275 / $26.375	500 ₽
GPT-5	$1.25 / $10	$1.319 / $10.55	—
Gemini 2.5 Pro	$1.25 / $5	$1.319 / $5.275	—

LiteAI does not provide non-Claude models — this is an architectural choice. If you need GPT/Gemini/Mistral, use OpenRouter.

---

2. Markups

This is the most nuanced section — markups can be explicit or hidden.

OpenRouter: explicit platform fee + hidden FX risk

Explicit markup:

• 5.5% PAYG fee — shown in the invoice, documented, appears in API response as usage.cost.
• 0–5% BYOK fee — incentivizes enterprise contracts.

Hidden Markup #1 — asymmetric I/O:

• Anthropic output is 5× pricier than input. OpenRouter inherits this ratio.
• If your use-case is agentic code (heavy input, low output), OpenRouter's effective markup might seem lower than a flat LiteAI rate.

Hidden Markup #2 — FX volatility:

• OpenRouter invoices in USD. At a 100 ₽/$ vs 80 ₽/$ exchange rate, your effective ruble-cost swings by ±25%.
• LiteAI uses a fixed RUB rate — your budget is predictable.

Hidden Markup #3 — minimum top-up:

• OpenRouter: $5 PAYG, $50 crypto. If you just want to test for $0.50, you'll have $4.50 stranded balance.
• LiteAI: Minimum package is 1M for 30 ₽ / $1.

LiteAI: markup "baked into" tariff

LiteAI does not publish procurement costs — we buy in bulk at enterprise rates. The markup is not zero, but it is:

• Fixed in RUB — FX immune.
• Uniform across models (except fable-5).
• Averages input/output — eliminates the 5× price gap.

LiteAI subsidizes lower-tier models (Haiku) via Opus-heavy volume, enabling a flat rate. For fable-5, LiteAI is roughly 4.5× cheaper than direct Anthropic procurement due to enterprise volume discounts.

Markup Conclusion

Scenario	OpenRouter effective markup	LiteAI effective markup
Claude Opus, agentic 80/20	~6%	flat 30 ₽/M, ~90% savings vs direct
Claude Sonnet, chatbot 30/70	~6%	flat 30 ₽/M, ~95% savings vs direct
Mixed Claude, 100M/mo	—	3,000 ₽ (predictable)

For RU-side teams, LiteAI wins on markup in every Claude scenario. OpenRouter wins only where model diversity is required.

---

3. Stability

Architecture

Parameter	OpenRouter	LiteAI
Backend	Multi-provider: Anthropic/Bedrock/Vertex/Azure	Single-backend: Anthropic direct via RU-infra
Auto-failover	✅ between providers	❌
Zero-data-log	Optional (data-policy routing)	✅ default (zero retention in LiteAI-infra)
Geographic routing	✅	❌
Status page	https://status.openrouter.ai/	Telegram channel + bot alerts
Contractual SLA	Enterprise	None

Uptime

OpenRouter maintains 99.9%+ uptime, but incidents occur (e.g., March 2024, when Bedrock fallback failed). Multi-provider reduces total failure risk, but does not eliminate it.

LiteAI claims 99.9% uptime. Single-backend means:

• If Anthropic direct is down — LiteAI is down.
• However: LiteAI's internal infra (Express/Nginx/Bifrost) is our responsibility. Our historical uptime over 12 months is >99.95%.
• Latency: adds ~30–80ms overhead. Usually negligible.

Regional stability

OpenRouter is hosted in US-East (primary). RU-side latency is 100–250ms, with risks of ISP/Cloudflare/AWS blocking.

LiteAI runs proxy nodes within Russia using direct peering. Latency is 30–80ms from Moscow, independent of international routing. This is critical for RU-side developers.

Conclusion on stability

For enterprise compliance/global SLA, OpenRouter wins (SLA, SOC2, audit logs). For RU-regional stability/low latency, LiteAI wins (RU-hosted, bypasses upstream routing instability).

---

4. Taxes

LiteAI wins decisively for RU-entities.

LiteAI: YooKassa, 54-FZ compliance

LiteAI uses YooKassa (RU-based). Receipts are generated via online cash register in accordance with 54-FZ.

• vat_code: 1 (no VAT/USN).
• Valid expense documentation for IP/OOO on USN.
• No currency control issues for RU-legal entities.

OpenRouter: Stripe, international

OpenRouter is a US company (Delaware).

• Stripe Invoice for enterprise, PAYG receipts are not 54-FZ compliant.
• Not a "primary document" for Russian accounting.
• Stripe blocks RU cards, requiring crypto-intermediaries for RU users.

---

Where each provider fails

LiteAI shortcomings

1. Model variety. 400+ (OR) vs only Claude (LiteAI).
2. Auto-failover. LiteAI has no multi-backend fallback.
3. Enterprise admin. No SSO, no audit logs export.
4. Non-RU support. No US Sales Tax support, strictly crypto-based for non-RU clients.

OpenRouter shortcomings

1. Claude price for RU. Fees + FX risks make it 9–47× pricier than LiteAI.
2. Payment. Stripe-blocked for RU.
3. VPN. OpenRouter requires foreign phone/card.
4. Predictability. Effective $/M floats with exchange rates.
5. Support. Discord vs human Telegram support.
6. Taxes. No 54-FZ fiscal receipts.

---

OpenRouter alternatives in Russia (closes the 430/mo query)

The "openrouter alternatives" search is the most relevant for RU developers, because OpenRouter is not directly payable from Russia. Below is a quick breakdown of alternatives covering different scenarios.

Provider	Models	RU payment	Claude Opus 4.8 (in/out per 1M)	Fiscal receipt	When to choose
OpenRouter (original)	400+	⚠️ crypto/VPN	$3 / $15 + 5.5% fee	❌	Need max model catalog
LiteAI	4 (Claude only)	✅ SBP/card/USDT	30 ₽ flat	✅	Claude-only production in RU
ProxyAPI	50+	✅ SBP/card	~$4 / $18 (~360 ₽/M)	✅	Multi-provider in one project
Bothub	30+	✅ card/crypto	~$3.5 / $16 (~315 ₽/M)	✅	Self-hosted / compliance
BotHub	20+	✅ card/SBP	~$4 / $17 (~360 ₽/M)	✅	No-code integrations

Honest conclusion: for pure Claude in Russia, LiteAI wins on price (9–12× cheaper than OpenRouter) and beats it on every non-monetary axis (taxes, latency, support). If you need GPT-5 + Gemini 2.5 Pro in one project, use ProxyAPI or Bothub as the RU-paid OpenRouter. For exotic open-source, stay on the original OpenRouter via crypto.

---

FAQ: top OpenRouter questions (covers the long tail)

How to get an OpenRouter API key from Russia?

Standard path: register at openrouter.ai, confirm email, create a key in Keys. RU problem: Stripe top-up is blocked for RU cards. Workarounds: crypto via Coinbase Commerce (min $50), Plati.ru/GGSel gift-cards (10–20% markup), or friend's foreign card. For Claude-only, get a LiteAI key for 30 ₽ via SBP — same sk-… format, no VPN.

How to top up OpenRouter from Russia?

пополнить openrouter — 1 027/mo. No direct path. Options: (1) Coinbase Commerce (USDT, min $50); (2) Plati.ru / GGSel gift-cards; (3) foreign card from a trusted contact; (4) for Claude-only, just use LiteAI 1M for 30 ₽.

Does OpenRouter work in Russia without VPN?

openrouter в россии — 601/mo. Partially. API endpoints usually work directly, but: registration may require foreign captcha, Stripe payments are 100% blocked, latency is 100–250 ms, and ISP-level blocking of AWS ranges is a real risk. For Claude-only LiteAI works fully in RU without VPN.

Is OpenRouter free?

openrouter бесплатно — 571/mo. Partially. Free-tier on some models (:free suffix: meta-llama/llama-3.3-70b:free, qwen/qwen-2.5-72b:free, etc.), but limits are ~20 req/min and ~50 req/day per model. For production-grade Claude access, use LiteAI 1M for 30 ₽ or Anthropic trial (if you have VPN).

What is OpenRouter API?

openrouter api — 2 490/mo. An OpenAI-compatible REST API at https://openrouter.ai/api/v1/chat/completions. Format is identical to OpenAI — any OpenAI SDK works, just swap base_url. Models use provider prefix: anthropic/claude-opus-4.8, openai/gpt-5, google/gemini-2.5-pro. The response includes usage.cost — actual USD cost including 5.5% fee.

Is OpenRouter a Russian service?

No. US company (OpenRouter Inc., Delaware), hosted in US-East, Stripe billing. If you need "OpenRouter but Russian", use LiteAI (Claude-only) or ProxyAPI (multi-provider). See the alternatives table above.

---

What changed since the last cut (2026-07-04 → 2026-07-22)

• OpenRouter grew from 29 446 to 38 978/mo (+32%). Weekly volume: 7 022 → 12 383 (+76% in two weeks).
• OpenRouter AI tracked as separate query: 13 092/mo — added a dedicated section.
• OpenRouter alternatives tracked: 430/mo — added a dedicated section.
• OpenRouter API 2 490/mo, OpenRouter key 1 542/mo, Пополнить OpenRouter 1 027/mo — all covered in the FAQ.
• Tariffs, fees, markups, and tax block unchanged from 22.07.2026.

---

Summary: LiteAI is the default for RU Claude users

If you are in Russia and working with Claude, LiteAI offers the best price-to-convenience ratio:

• Tariffs: 30 ₽/M flat, RUB-denominated.
• Markup: Negligible/negative effective markup via enterprise bulk contracts.
• Stability: 99.9% uptime, RU-infra, 30–80ms latency, immune to upstream ISP blocks.
• Taxes: 54-FZ fiscal receipts for RU-legal entities.

OpenRouter remains a secondary tool for those requiring a broader model catalog, multi-provider routing, or US/EU enterprise compliance.

Start with the 1M for 30 ₽ package at liteai.tech/pricing — enough for 20–30 Opus 4.8 test sessions to decide if you need OpenRouter's model variety.