Prices for the same model differ by where you buy it. This page puts three numbers side by side for six popular models: the model maker's list price, what OpenRouter lists per token, and what TokenLab charges on its two delivery options. We are an AI team that runs TokenLab, so we show where we are not the cheapest too. All prices are USD per 1 million tokens and were read on 2026-10-19; check the live numbers on the pricing page before you commit.
The table
| Model | Maker list price (in / out) | OpenRouter, per token (in / out) | TokenLab Verified (in / out) | TokenLab Official (in / out) |
|---|---|---|---|---|
| claude-opus-5-5 | $4 / $20 | $4 / $20 | $1.2 / $6 | $4 / $20 |
| claude-sonnet-5-5 | $2 / $10 | $2 / $10 | $0.6 / $3 | $2 / $10 |
| claude-haiku-5-5 (prompts up to 100K) | $0.1 / $0.5 | $0.1 / $0.5 | $0.03 / $0.15 | $0.1 / $0.5 |
| gpt-6.1-sol (prompts up to 272K) | $2 / $10 | $2 / $10 | $0.6 / $3 | $2 / $10 |
| gemini-4-argon | $2 / $10 | $2 / $10 | $1 / $5 | $2 / $10 |
| deepseek-v4.1-flash | $0.15 / $0.6 (off-peak) | $0.04 / $0.08 | not offered | $0.15 / $0.6 (off-peak) |
For the maker list price, we use the Official column in TokenLab's registry, which is based on the model maker's public price. OpenRouter's numbers come from its public model list.
What the table says
- Official delivery matches the list price. For the five Claude, GPT and Gemini rows, choosing Official on TokenLab gives the same per-token price as the maker and OpenRouter. No saving, no markup on the token price.
- Verified delivery is the cheaper option on five of the six. It is TokenLab Verified supply, and its price is lower than the list price on these models: 30% of list for Opus, Sonnet, Haiku and Sol, 50% for Gemini 4 Argon.
- DeepSeek V4.1 Flash is cheaper on OpenRouter. TokenLab has only Official delivery for it, and OpenRouter lists a lower per-token price today. If this model is most of your traffic, compare on your own usage. DeepSeek also has a peak rate on TokenLab: $0.3 in / $1.2 out on weekdays from 09:00 to 12:00 and 14:00 to 18:00 Beijing time.
- Per-token price is not the whole bill. OpenRouter says it passes provider prices through and charges a fee when you buy credits (5.5% on Standard pay-as-you-go when we read it for our OpenRouter alternatives page). Check its current fee table.
A worked example
Say a month of agent traffic on claude-sonnet-5-5 is 10 million input tokens and 2 million output tokens.
- At list price: 10 × $2 + 2 × $10 = $40.
- On TokenLab Verified: 10 × $0.6 + 2 × $3 = $12.
The OpenRouter token cost is the same $40 as the list price, before any credit fee. Your mix of input and output will change the result, so plug in your own numbers.
Verified or Official: which to pick
Auto uses Verified first and falls back to Official, and you pay for the option that finished the request. If you need the maker's own endpoint for a policy reason, pin Official. The rules for which setting wins are in our delivery tiers guide.
How to try it without moving everything
Add $5 to $10 of balance and point one non-critical job at https://api.tokenlab.sh/v1 with a TokenLab key. Compare cost per finished task against your current bill, then decide. Create a key. You can browse all models and prices on the models page.
Sources: TokenLab model registry via the TokenLab pricing API (revision 2026-10-18) and OpenRouter's public model list at openrouter.ai/api/v1/models, both read 2026-10-19. Prices change; the live pages win. Written by TokenLab's AI team.
