Home → Compare → APICLAN vs paying the provider directly
The comparison people expect is a price table. The price table is the easy part. The part worth reading is what the lower price does not include.
| Per 1M tokens | Official list | Here | You save |
|---|---|---|---|
claude-opus-5-5 | $4.00 / $20.00 | $0.900 / $4.50 | 78% |
claude-opus-5 | $5.00 / $25.00 | $1.12 / $5.62 | 78% |
claude-sonnet-5 | $2.00 / $10.00 | $0.450 / $2.25 | 78% |
gpt-6-astra | $10.00 / $50.00 | $2.00 / $10.00 | 80% |
gpt-5.6-sol | $5.00 / $30.00 | $1.00 / $6.00 | 80% |
grok-4.7 | $2.00 / $6.00 | $0.400 / $1.20 | 80% |
deepseek-v4-pro | $1.32 / $3.96 | $0.675 / $2.02 | 49% |
kimi-k3 | $2.95 / $14.74 | $1.12 / $5.60 | 62% |
glm-5.3 | $1.40 / $4.40 | $0.280 / $0.880 | 80% |
These are published rates, not introductory offers, and they are the same numbers your invoice uses. Each model's page shows the arithmetic.
A price table is easy to write and impossible for a reader to verify. So here is the check we run on ourselves: for every model with real traffic on this account, we take the provider's published rate and compare it against what we were actually billed per million tokens, derived from the request log rather than from a marketing page.
| Per 1M tokens | Published rate | Our invoices | Agreement |
|---|---|---|---|
gpt-6-astra | $10.00 / $50.00 | $11.14 / $50.77 | +11% — tiered pricing |
gpt-6-sol | $2.00 / $10.00 | $2.00 / $10.00 | matches |
gpt-6.1-sol | $2.00 / $10.00 | $2.16 / $10.03 | +8% — tiered pricing |
gpt-5.6-sol | $5.00 / $30.00 | $5.00 / $30.00 | matches |
gpt-5.6-terra | $2.00 / $12.00 | $2.00 / $12.00 | matches |
gpt-5.6-luna | $0.20 / $1.20 | $0.20 / $1.20 | matches |
gpt-5.5 | $5.00 / $30.00 | $5.00 / $30.00 | matches |
grok-4.7 | $2.00 / $6.00 | $2.00 / $6.00 | matches |
grok-4.6 | $2.00 / $6.00 | $2.00 / $6.00 | matches |
claude-fable-5-1 | $10.00 / $50.00 | $10.00 / $50.00 | matches |
claude-fable-5 | $10.00 / $50.00 | $10.00 / $50.00 | matches |
claude-opus-5-5 | $4.00 / $20.00 | $4.00 / $20.00 | matches |
claude-opus-5 | $5.00 / $25.00 | $5.00 / $25.00 | matches |
claude-sonnet-5 | $2.00 / $10.00 | $2.00 / $10.00 | matches |
deepseek-v4-pro | $1.32 / $3.96 | $0.59 / $3.93 | +-55% — tiered pricing |
deepseek-v4-flash | $0.44 / $1.32 | $0.80 / $1.85 | +82% — tiered pricing |
deepseek-v4-flash-vision-exp | $0.44 / $1.32 | $0.15 / $0.60 | +-66% — tiered pricing |
glm-5.3 | $1.40 / $4.40 | $1.40 / $4.40 | matches |
kimi-k3 | $2.95 / $14.74 | $2.96 / $15.35 | matches |
gemini-4-argon | $2.00 / $10.00 | $1.00 / $4.99 | +-50% — tiered pricing |
gpt-5.6 | $4.00 / $20.00 | $5.00 / $30.00 | +25% — tiered pricing |
gpt-5.4-mini | $0.75 / $4.50 | $0.75 / $4.50 | matches |
claude-opus-4-6 | $5.00 / $25.00 | $5.00 / $25.00 | matches |
claude-sonnet-4-6 | $3.00 / $15.00 | $3.00 / $15.00 | matches |
17 of 24 models bill exactly at the published rate. Checked 2026-10-20.
Where a model does not match, the cause is tiered pricing rather than a billing error,
and it is worth knowing before you budget. gpt-6-astra reprices a request
whose input exceeds 272,000 tokens — the whole call, not just the tokens past the
threshold. Long agent runs cross that line routinely, which is why its observed average
sits above the headline rate even though every individual call was billed correctly.
This is also why the prices on this site are the published rates rather than our observed averages: an average that quietly includes a long-context surcharge would make our own discount look smaller than it is, and would misstate the provider's list price to anyone reading the page.
If you are shipping a product with real customers whose data you are answerable for, or if your company needs a signed agreement before it can use a vendor, go direct. The price difference will not cover the first compliance conversation you cannot win.
If you are building, prototyping, running agents against your own code, or operating a small product where the API bill is a line item you feel every month, the arithmetic above is the argument, and there is nothing to negotiate to get it.
A third option people forget: use both. Direct for the customer-facing path that needs the contract, and a relay for development, evaluation and internal tooling, which is usually where most of the token count lives and none of the compliance risk does.
No subscription, no monthly minimum, no sales call. Top up with USDT and spend what you use — 1 USDT gives you 2 credits of API balance.
Read the 30-second quickstartCompetitor terms quoted from their own published documentation and checked on 2026-10-20. Our prices are regenerated automatically from live billing data.