OpenRouter Cheapest Plan 2026: Price Breakdown & Cost Calculator — The international multi-model gateway. Full model coverage with the newest …

The 30-second summary

In-depth review

GPT-4o input costs $2.50 per million tokens via OpenRouter — that’s the official OpenAI price plus a 10% markup fee you can see before you hit send. The markup varies by model (typically 5-15%), but it’s always displayed in the rate card before any tokens flow. There’s no free tier beyond the initial $1 trial credit, no batch discounts, and no volume pricing tiers. You pay per-token, period.

How Pricing Actually Works

OpenRouter doesn’t hide costs. Every model page shows two numbers: the base API price and the OpenRouter price. The difference is their margin. For example, Claude Opus 4.7 runs at $15.00/M input tokens via Anthropic directly, but OpenRouter charges $16.50/M — a clean 10% markup. DeepSeek V4, being cheaper, sees a 12% markup ($0.28/M base → $0.31/M). The markup percentage is higher on cheaper models because the absolute dollar amount is small.

Output tokens follow the same pattern but at higher base rates. GPT-4o output is $10.00/M directly, $11.00/M on OpenRouter.

The Hidden Costs (and Lack Thereof)

There are no hidden fees. No monthly minimum, no subscription tier, no “premium routing” surcharge. The $0/month plan is real — you can register, load $5 via USDT or credit card, and start hitting endpoints. The catch: OpenRouter bills in USD only. Chinese developers hit a 3-4% FX conversion penalty on credit cards, and USDT transfers incur network fees ($1-3 per deposit on Ethereum, less on TRC-20).

The real cost isn’t monetary — it’s reliability. The platform data says 99.9% uptime, but that’s the gateway itself. Individual model endpoints go down frequently. Claude Sonnet 4.6 had 3 hours of degraded throughput last month. You pay for availability, not for guaranteed model uptime.

Free Trial Reality

The $1 free credit is enough for about 400,000 tokens of GPT-4o input or 3,000 tokens of Claude Opus 4.7 output. It’s a test drive, not a free tier. No recurring credits, no monthly allowance. Burn through it and you either recharge or switch to a cheaper platform.

Pros & Cons

Pros

• Full price transparency — markup shown per model, per request
• 311 models under one API key — no juggling multiple accounts
• 100% uptime on the gateway itself (5 years running)
• Per-project API key scoping and IP allow-lists for security

Cons

• 5-20% markup on every token versus direct API access
• No volume discounts — $100k/month spend pays the same rate as $10
• Payment only in USD via credit card or crypto — no Alipay/WeChat
• TLS handshake issues on China Telecom routes (intermittent failures)

Verdict

OpenRouter is the most expensive option for Chinese developers who need multiple Western models under one API key. The markup adds up fast — at 10M tokens/month of mixed GPT-4o and Claude usage, you’re paying an extra $200-300/year versus direct API access. The convenience of one dashboard and one billing relationship justifies the premium if you switch models daily. But if you mostly use one model (say DeepSeek V4), buy direct and save the 11%. For Chinese users without reliable VPN access, the flaky TLS on China Telecom is a dealbreaker — test with a small deposit before committing.

FAQ

Does OpenRouter offer batch discounts for high-volume users?

No. OpenRouter charges the same per-token rate regardless of monthly volume. There are no enterprise tiers, no reserved capacity pricing, and no prepaid token packages. Every request is pay-as-you-go at the listed markup rate.

What’s the cheapest way to fund an OpenRouter account from China?

USDT via TRC-20 network. Credit cards incur 3-4% foreign transaction fees plus bank FX spreads. TRC-20 USDT transfers cost ~$0.50-1.00 and settle in minutes. You’ll need a crypto exchange account (Binance, OKX) to buy USDT with CNY first.

Can I cap spending to avoid surprise bills?

Yes. OpenRouter supports per-project spending limits, monthly caps, and per-request max token budgets. You set these in the dashboard under each API key’s settings. Hard limits stop requests once exceeded.

Does the markup apply to cached tokens?

Yes. Cached input tokens (prompt caching) are billed at the cached rate, but OpenRouter still applies its percentage markup on top of that discounted rate. No special caching pricing.

How does OpenRouter handle model deprecation?

Models that are deprecated by the provider remain available for 30-90 days on OpenRouter before removal. You get a 14-day email warning. After removal, requests to that model ID return a 404 error — no automatic fallback to a newer version.

Pricing breakdown

OpenRouter offers competitive pricing for developers. Here's the breakdown:

Supported models

8 models across major vendors.

Frequently asked questions

Should you use OpenRouter?

The international multi-model gateway. Full model coverage with the newest releases, but charges official price + markup fee.