Frontier and open-weight models.Up to 63% below API list price.Zero data retention.*
The same models you already use, through a drop-in, OpenAI-compatible API, priced by a published tide that only goes down until midnight UTC.
- ZDR on Private models
- One base URL
- Pay by card
- Price locked per request
ZDR guaranteed on Private models. Tokentide never stores prompts or outputs, on any model.
Tide levelStandard
Standard multiplier
0.700×−30.0%below API list price
- Low Tide
- 0.490×
- Tomorrow opens at
- 0.750×
Today’s published tide
- Standard
- Low Tide
Now
Live prices
Same models. Today’s price.
Tokentide nowAPI list priceRefreshed every 15 seconds
All models and pricesUSD per 1M tokens · At the current Standard multiplier (0.700×). Low Tide is cheaper while it’s open.
Drop-in
Change one line. Keep everything else.
Tokentide speaks the OpenAI API. Point your SDK at our base URL, use your Tokentide key, and your code, prompts and streaming keep working.
- OpenAI SDKs for Python and TypeScript
- Vercel AI SDK, LangChain and LlamaIndex
- Venice-style /api/v1 paths on every host
- Streaming, tools, vision, images and audio
import osfrom openai import OpenAI client = OpenAI(Removed: base_url="https://api.openai.com/v1",Added: base_url="https://api.tokentide.app/v1", api_key=os.environ["API_KEY"], # your tt_live_… key) reply = client.chat.completions.create( model="zai-org-glm-5-2", messages=[{"role": "user", "content": "Summarize this contract."}],)print(reply.choices[0].message.content)Why is it cheaper?
Capacity that would expire at midnight, sold before it does.
- 1$1 / DIEM / day
Step 1: Stakers get a daily allowance
Each staked DIEM grants $1 of Venice API credit per day, at Venice’s normal prices.
- 200:00 UTC
Step 2: Unused allowance expires at 00:00 UTC
Credit doesn’t roll over. Whatever isn’t used by midnight UTC is gone for good.
- 3api.tokentide.app
Step 3: Tokentide sells it first
We resell that spare capacity through a private API, so you pay less and nothing goes to waste.
Venice remains the source of truth for its models. Tokentide is an independent service, not affiliated with Venice.ai.
Private by design
Zero data retention, guaranteed on Private models.
Your prompts and outputs stream through memory and are never written to disk, logs or analytics. On Private models the inference provider retains nothing either, so retention is zero end to end.
Data stored in the United States
Metadata and the ledger live in one database in AWS US East.
Private-only keys
Lock an API key to Private models, so nothing it sends reaches a model without zero data retention.
Never stored
- Prompts and messages
- Files, images and audio
- Model outputs
- Your identity, toward model providers
Stored, metadata only
- Account and billing records
- Request metadata: model, tokens, cost (90 days)
- IP addresses as daily-salted hashes (30 days)
- Security audit log (2 years)
Privacy tiers, labeled on every model
- Private
Run with contractual zero data retention.
- TEE
Runs inside hardware enclaves (trusted execution environments).
- E2EE
Encrypted on your device to an attested enclave key. We only relay ciphertext.
- AnonymizedOpt-in
Third-party models such as Claude, GPT and Gemini, sent without your identity. Opt-in: the provider sees content under its own policy.
Fair by design
A tide table, not a slot machine.
Every day we publish the price curve in advance. It falls gently through the UTC day, and can only move up, within published bounds, when capacity gets tight.
The Standard multiplier falls from 0.720× at 00:00 UTC to 0.504× at 24:00 UTC.
The price only goes down during the day
The published table never rises. Short, bounded surges protect capacity when demand spikes.
Locked when your request starts
You pay the multiplier at admission, even if the tide moves mid-stream. Every response tells you which one.
Low Tide for flexible work
Opt in per request for non-streaming jobs and pay up to 63% less while spare capacity is open, usually in the evening UTC.
No hidden fees
No subscriptions, seats or surcharges. You pay the API list price times the multiplier, nothing else.
Pay your way
Pay by card. Top up in seconds.
Every top-up funds the same USD balance and sees the same market price.
Card, worldwide
Top up from $10 with Stripe Checkout. The VAT, GST or sales tax for your location is added at checkout, and you get a receipt and an invoice by e-mail.
- Apple Pay, Google Pay and Link
- Receipts and invoices from Stripe
- No surcharges
Base URL
One base URL. Every model.
Point any OpenAI-compatible SDK at it. Tokentide runs in the United States (AWS US East) and keeps what it stores there; the inference provider decides where models run.
https://api.tokentide.app/v1Comparison
Tokentide or the direct API?
Same models, same privacy tiers. Here’s what changes.
| Feature | Tokentide | Direct API |
|---|---|---|
| Price | 25–63% below API list price, live | API list price |
| Pricing transparency | Published tide tables, price locked per request, multiplier on every response | Published price sheet |
| Models | Frontier and open-weight text models, plus image, video, music, audio and decision models | Full catalog, including voice cloning |
| Data retention | No content stored; metadata only, in the US | Provider’s privacy policy; zero retention on Private models |
| API compatibility | OpenAI- and Venice-compatible | OpenAI-compatible |
| Discount lane for batch work | Low Tide: opt-in, up to 63% off | — |
Direct API means Venice’s own API, based on its public documentation as of October 2026. Venice and DIEM are trademarks of their respective owners.
Savings calculator
What would you pay?
Estimate a month at Tokentide prices from what you spend today at API list prices, and when your traffic runs.
$50$50K
Non-streaming jobs that can wait for spare capacity
You save
$757per month
At Tokentide
$1,242.83
Effective multiplier
0.621×
≈ 1.5B tokens per month with this mix
Estimate based on today’s published tide table. Actual prices depend on when each request starts and on Low Tide availability; surges can raise them within published bounds.
Trust
Built to be checked.
Live status
Every endpoint family, with the upstream provider shown separately.
View statusSecurity practices
Passkeys and two-factor sign-in, strict CSP, HSTS preload, encrypted secrets and a hash-chained audit log.
Security overview- Quarterly
Transparency report
Every quarter: uptime, how the market behaved, and the metadata requests we received. We hold no content to disclose.
Independent
Tokentide is an independent service, not affiliated with Venice.ai. Venice remains the source of truth for its models.
FAQ
Questions, answered.
Is it really the same model?
Yes. Your request goes to Venice’s API with the model you name; we never swap models silently. Model ids, capabilities and list prices come straight from Venice’s catalog.
Why is it cheaper?
Staked DIEM grants a daily Venice allowance that expires at 00:00 UTC. We sell the part that would go unused before it expires, at a published discount.
What does zero data retention cover?
Tokentide never stores prompts, files, images, audio or outputs, on any model. On Private models Venice retains nothing either, so retention is zero end to end. Anonymized third-party models are opt-in; their providers may retain content under their own policies, without your identity.
How is the price set?
A published tide table sets the Standard multiplier for every minute of the UTC day, from about 25% below API list price in the morning to about 47% below at midnight. Bounded surges can raise it when capacity is tight. The multiplier is locked when your request starts and returned in the X-Tokentide-Multiplier header.
What is Low Tide?
An opt-in, best-effort lane for non-streaming work that sells only spare capacity, up to 63% below API list price. When it’s closed you get a 409 with the forecast opening time, or you can fall back to Standard automatically.
What happens at midnight UTC?
That’s slack water: the daily allowance resets. A request that might still be running after the reset is priced at the next day’s published opening price, so nothing changes mid-request.
How can I pay?
By card through Stripe Checkout from $10 (the tax for your location is added at checkout). Apple Pay, Google Pay and Link work too.
Do my credits expire?
We never confiscate paid credit. Promotional and bonus credit expires after 12 months. You can ask for a refund of unused paid credit within 30 days, and EU consumers have a 14-day withdrawal right.
Where is my data processed?
Tokentide runs in the United States (AWS US East), and the metadata and billing records we store stay there. Request content is never stored. Venice decides where inference runs.
Are you affiliated with Venice?
No. Tokentide is an independent service, not affiliated with Venice.ai.
Catch tonight’s tide.
Create a key in under a minute, change one line, and pay less for the same private models.