Pricing · Taifoon Cloud
Resold prices
Taifoon resells models from other providers. For each model you get one route: the cheapest Taifoon sells now. You pay that provider’s own price plus one cut of 20 %, and nothing else. Every price below can be checked against the provider’s own sheet.
1. How a resold price is made
Taifoon buys each call from the provider at the provider’s published price. It adds one cut of 20 % over that price, rounded up to the micro-USDC, and that sum is the price you pay. The search, the gateway and every channel quote the same price: there is no second cut.
What you pay leaves again in two named legs, and they add up to your payment exactly:
| Leg | Goes to | How much |
|---|---|---|
seller_share | the provider that served the call | its own published price |
reseller_markup | Taifoon | the cut, 20 % over the provider’s price |
A channel that takes its own fee at the source, such as an agent marketplace, adds that fee as a third leg of its own. It is shown on the quote and is never part of Taifoon’s leg.
One minimum applies: a single x402 payment is at least 0.0012 USDC, because smaller payments are refused by the payment network. A call priced below that is charged the minimum, and the difference is Taifoon’s. Paying from a key’s balance has no minimum.
2. Who allows each resale
Taifoon publishes prices only for these providers. Their terms, or their written permission, allow resale. A provider still waiting on a reseller agreement is not on the list.
| Provider | What allows it | Price sheet |
|---|---|---|
| Akash Network | permissionless leases: no resale restriction found in the network terms; a provider runs our container, we serve our endpoint | console-api.akash.network/v1/gpu-prices |
| DeepInfra | You are allowed to serve our models through your gateway. (DeepInfra, 2026-10-28, the express permission its terms 11(a)(viii) ask for) | api.deepinfra.com/models/list |
| NEAR AI Cloud | NEAR AI's written yes of 2026-10-22: it supports routing inference through our gateway, x402 payments and prepaid credits included; ToS §4.2(x) alone forbids resale; Chutes routes follow Chutes' terms (§7.1) | cloud-api.near.ai/v1/models |
| Venice AI | ToS 7.3: a reseller serves end users bound by terms as protective and is primarily liable; resale allowed by the owner 2026-11-03 (_VENICE_SELLS_v1_); the partner discount stays owner step venice-reseller-agreement | api.venice.ai/api/v1/models?type=text |
3. Check a price yourself
Three public calls, no key needed. The first names, for every resold model, the provider’s price, the cut and where the price was read. The second is the provider’s own sheet. The third shows the legs of a quote.
# what you pay, what the provider charges, the cut, and where its price was read
curl -s https://coord.taifoon.dev/v1/models | jq '.data[] | select(.taifoon.resold) | {id, owned_by, pay: .taifoon.usd_per_1m, resold: .taifoon.resold}'
# the provider's own sheet (DeepInfra: cents per token × 10,000 = USD per 1M tokens)
curl -s https://api.deepinfra.com/models/list | jq '.[] | select(.model_name == "meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo") | .pricing'
# the legs of one quote: the provider's share, the cut, the sum
curl -s 'https://coord.taifoon.dev/v1/match/search?tag=llama-3.1-8b' | jq '.hits[] | {venue, legs: .legs.legs}'4. The price list
One route per model: the one on sale now, and of those the cheapest. USD per 1 million tokens, input / output. Under each model: the day its price was read from the provider’s sheet, and how you can tell which model answered (TEE-signed: the answer is signed by a sealed machine). “Not on sale yet” means no provider on this list sells that model yet: the price is set, but Taifoon has not yet opened or funded its account there.
| Model | Served by | Provider price | You pay |
|---|---|---|---|
deepseek-r1-0528read 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.5 / 2.15 | 0.6 / 2.58 |
deepseek-v3.1read 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.25 / 0.95 | 0.3 / 1.14 |
deepseek-v3.2read 2026-10-16 · model named by the provider · on sale | DeepInfra | 0.26 / 0.38 | 0.312 / 0.456 |
deepseek-v4-flashread 2026-10-19 · model named by the provider · on sale | DeepInfra | 0.09 / 0.18 | 0.108 / 0.216 |
deepseek-v4-flash-0731read 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.06 / 0.18 | 0.072 / 0.216 |
deepseek-v4-flash-vision-expread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.44 / 1.32 | 0.528 / 1.584 |
deepseek-v4-proread 2026-10-28 · model named by the provider · on sale | DeepInfra | 1.3 / 2.6 | 1.56 / 3.12 |
deepseek-v4-pro-0813read 2026-10-28 · model named by the provider · on sale | DeepInfra | 1.3 / 2.6 | 1.56 / 3.12 |
deepseek-v4.1-flashread 2026-10-27 · model named by the provider · on sale | DeepInfra | 0.2 / 0.6 | 0.24 / 0.72 |
deepseek-v4.1-flash-1017read 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.2 / 0.6 | 0.24 / 0.72 |
gemma-3-12b-itread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.05 / 0.15 | 0.06 / 0.18 |
gemma-3-27b-itread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.08 / 0.16 | 0.096 / 0.192 |
gemma-3-4b-itread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.05 / 0.1 | 0.06 / 0.12 |
gemma-4-26b-a4b-itread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.07 / 0.34 | 0.084 / 0.408 |
gemma-4-26b-a4b-it-hereticread 2026-10-28 · model named by the provider · on sale | Venice AI | 0.19 / 0.88 | 0.228 / 1.056 |
gemma-4-31b-itread 2026-10-28 · model named by the provider · on sale | Venice AI | 0.12 / 0.36 | 0.144 / 0.432 |
glm-4.6read 2026-10-28 · model named by the provider · on sale | Venice AI | 0.43 / 1.75 | 0.516 / 2.1 |
glm-4.7read 2026-10-19 · model named by the provider · on sale | DeepInfra | 0.4 / 1.75 | 0.48 / 2.1 |
glm-4.7-flash-hereticread 2026-10-28 · model named by the provider · on sale | Venice AI | 0.07 / 0.4 | 0.084 / 0.48 |
glm-5read 2026-10-28 · model named by the provider · on sale | Venice AI | 1 / 3.2 | 1.2 / 3.84 |
glm-5.1read 2026-10-28 · model named by the provider · on sale | Venice AI | 1.54 / 4.84 | 1.848 / 5.808 |
glm-5.2read 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.75 / 2.4 | 0.9 / 2.88 |
glm-5.3read 2026-10-27 · model named by the provider · on sale | DeepInfra | 0.9 / 4 | 1.08 / 4.8 |
glm-5.3-flashread 2026-10-22 · TEE-signed · on sale | NEAR AI Cloud | 0.08 / 0.5 | 0.096 / 0.6 |
gpt-oss-120bread 2026-10-19 · model named by the provider · on sale | DeepInfra | 0.037 / 0.17 | 0.0444 / 0.204 |
gpt-oss-20bread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.03 / 0.14 | 0.036 / 0.168 |
granite-4.2-8bread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.06 / 0.25 | 0.072 / 0.3 |
hy4-previewread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.834 / 2.501 | 1.0008 / 3.0012 |
inkling-smallread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.45 / 1.2 | 0.54 / 1.44 |
kimi-k2.5read 2026-10-28 · model named by the provider · on sale | Venice AI | 0.56 / 3.5 | 0.672 / 4.2 |
kimi-k2.6read 2026-10-27 · model named by the provider · on sale | DeepInfra | 0.75 / 3.5 | 0.9 / 4.2 |
kimi-k2.7-coderead 2026-10-28 · model named by the provider · on sale | Venice AI | 0.75 / 3.5 | 0.9 / 4.2 |
kimi-k3read 2026-10-27 · model named by the provider · on sale | DeepInfra | 2.85 / 14.25 | 3.42 / 17.1 |
l3-8b-lunaris-v1read 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.04 / 0.05 | 0.048 / 0.06 |
ling-3.0-flashread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.06 / 0.18 | 0.072 / 0.216 |
ling-3.0-flash-vlread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.06 / 0.18 | 0.072 / 0.216 |
ling-3.1-flashread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.3 / 0.9 | 0.36 / 1.08 |
llama-3.1-8b-instructread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.02 / 0.04 | 0.024 / 0.048 |
llama-3.2-3bread 2026-10-28 · model named by the provider · on sale | Venice AI | 0.15 / 0.6 | 0.18 / 0.72 |
llama-3.3-70b-instructread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.1 / 0.32 | 0.12 / 0.384 |
llama-4-scout-17b-16e-instructread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.1 / 0.3 | 0.12 / 0.36 |
llama-guard-4-12bread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.18 / 0.18 | 0.216 / 0.216 |
mimo-v2.5read 2026-10-28 · model named by the provider · on sale | Venice AI | 0.4 / 2 | 0.48 / 2.4 |
mimo-v2.6-flash-rlread 2026-10-28 · model named by the provider · on sale | Venice AI | 0.175 / 0.35 | 0.21 / 0.42 |
minimax-m2.5read 2026-10-28 · model named by the provider · on sale | Venice AI | 0.27 / 0.95 | 0.324 / 1.14 |
minimax-m2.7read 2026-10-28 · model named by the provider · on sale | Venice AI | 0.375 / 1.5 | 0.45 / 1.8 |
minimax-m3read 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.28 / 1.1 | 0.336 / 1.32 |
mistral-small-24b-instruct-2501read 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.05 / 0.08 | 0.06 / 0.096 |
mistral-small-3.2-24b-instruct-2506read 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.075 / 0.2 | 0.09 / 0.24 |
muse-glimmer-30bread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.3 / 1.2 | 0.36 / 1.44 |
muse-spark-open-816b-a42bread 2026-10-28 · model named by the provider · on sale | Venice AI | 1.5625 / 5.3125 | 1.875 / 6.375 |
nemotron-3-nano-4bread 2026-10-19 · our container · on sale | Taifoon GPU (our own) | our own price | 0.04 / 0.04 |
phi-4read 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.07 / 0.14 | 0.084 / 0.168 |
qwen2.5-72b-instructread 2026-10-16 · model named by the provider · on sale | DeepInfra | 0.36 / 0.4 | 0.432 / 0.48 |
qwen2.5-7b-instructread 2026-10-28 · model named by the provider · on sale | Venice AI | 0.05 / 0.13 | 0.06 / 0.156 |
qwen3-14bread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.12 / 0.24 | 0.144 / 0.288 |
qwen3-235b-a22b-instruct-2507read 2026-10-19 · model named by the provider · on sale | DeepInfra | 0.09 / 0.55 | 0.108 / 0.66 |
qwen3-235b-a22b-thinking-2507read 2026-10-28 · model named by the provider · on sale | Venice AI | 0.45 / 3.5 | 0.54 / 4.2 |
qwen3-30b-a3bread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.12 / 0.5 | 0.144 / 0.6 |
qwen3-coder-480b-a35b-instructread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.3 / 1 | 0.36 / 1.2 |
qwen3-next-80b-a3b-instructread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.09 / 1.1 | 0.108 / 1.32 |
qwen3-vl-235b-a22b-instructread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.2 / 0.88 | 0.24 / 1.056 |
qwen3-vl-30b-a3b-instructread 2026-10-22 · TEE-signed · on sale | NEAR AI Cloud | 0.15 / 0.55 | 0.18 / 0.66 |
qwen3.5-27bread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.26 / 2.6 | 0.312 / 3.12 |
qwen3.5-35b-a3bread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.14 / 1 | 0.168 / 1.2 |
qwen3.5-397b-a17bread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.45 / 3 | 0.54 / 3.6 |
qwen3.5-9bread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.1 / 0.15 | 0.12 / 0.18 |
qwen3.6-27bread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.32 / 3.2 | 0.384 / 3.84 |
qwen3.6-35b-a3bread 2026-10-28 · model named by the provider · on sale | DeepInfra | 0.1 / 0.95 | 0.12 / 1.14 |
qwen3.8-2.4t-a95bread 2026-10-28 · model named by the provider · on sale | Venice AI | 2.5 / 7.5 | 3 / 9 |
qwen3.8-27bread 2026-10-22 · TEE-signed · on sale | NEAR AI Cloud | 0.04 / 1.35 | 0.048 / 1.62 |
Live availability per model, on this deployment, is https://coord.taifoon.dev/v1/models (taifoon.available).
5. Partners: your own markup
A partner, for example one building on n8n, resells under its own key. For each customer it creates a child with its own markup. The customer pays the price above plus that markup. The markup is the partner’s own leg on every call. Taifoon’s leg stays the same and is never raised by a partner’s markup.
POST https://coord.taifoon.dev/v1/tenant/keys
{ "action": "child", "label": "customer-a", "markup_bps": 500, "cap_day_usdc": 5 }| Partner | Markup it may set | Set by |
|---|---|---|
| self-serve (any key) | 0 % to 500 % per child | the partner, when it creates the child |
| enrolled partner | 0 % to 100 % | its enrolment with Taifoon |
Each child’s calls, what it owes and the partner’s margin: GET https://coord.taifoon.dev/v1/access/reseller. Revoke or re-cap a child with POST https://coord.taifoon.dev/v1/tenant/keys.