How I would run the weekly operating loop around Venice's catalog, channels and DIEM, and a working desk that reads it all live from public data.
Venice sells private and uncensored AI three ways: a consumer app with Pro plans, a pay-per-token API, and DIEM, a token that carries $1 of API credit per day. The same catalog of 382 models sits behind all three.
Keep that catalog, its prices and its channels consistent and announced: launches, removals, channel listings, provider uptime, and the DIEM numbers holders and builders plan around. Most of it is a weekly loop that can be written down and automated.
A desk on eight public sources: the catalog, docs, status page, OpenRouter, Base contracts, DefiLlama and Dexscreener. It checks that Venice's own sources agree, prices every channel, reads DIEM and VVV straight from Base, tracks catalog changes and writes the weekly brief. 74 rule tests run in the visitor's browser.
| Takeaway | Why it matters |
|---|---|
| One catalog, several price lists | Direct API, OpenRouter, Pro plan API allowances and DIEM all quote the same models. Today they agree: all 35 shared models match on list price, 7 with a temporary OpenRouter discount. Keeping it that way is a weekly job. |
| Removals are announced, then enforced | Three models leave by 30 October, published on the deprecations page with emails to recent callers. autoRemap is off on all three, so the follow-through is tracking each old id to zero. |
| OpenRouter is a second storefront | Venice is a provider on 43 matched models there. Price and uptime decide routing share in that channel, measured publicly every day. |
| DIEM is a real cost line | About 28,300 DIEM are staked, so roughly $28,000 a day of inference is drawn as credit rather than cash. At about $1,650 a DIEM the market prices that credit at a ~22% yearly yield. Both numbers belong in the weekly read. |
Founded by Erik Voorhees, launched May 2024. Private by design: chats stay in the user's browser, and Venice-run models keep no prompt data. VVV launched January 2025 and DIEM August 2025. In July 2026 Venice raised a $65M Series A at a $1B valuation led by Dragonfly, its first outside capital, and reported 3.5M registered users and 1.3T tokens a month; its pages now say over 4 million users.
| Line | What it is | How it earns |
|---|---|---|
| Consumer app | Chat, image, video, music and characters on venice.ai and mobile. Free tier, Pro at $18 a month, Pro Plus at $68 with $75 of monthly API spend, Max at $200 with $225. | Subscriptions; credits for video and upscales (100 credits = $1) |
| API | OpenAI-compatible at api.venice.ai: 129 text, 43 image, 138 video, 18 music, 11 TTS, 9 embedding and 5 speech-to-text models today. | Pay per token or per job at catalog prices |
| DIEM | Mint by locking staked VVV. Each staked DIEM gives $1 of API credit per day, without end. Tradable on Base. | Turns staked VVV into prepaid API capacity |
| VVV | Base token; staking earns emissions (2M VVV a year from 1 October) and is the input for minting DIEM. | $5 of every $100 of credit purchases buys and burns VVV |
| OpenRouter provider | Venice serves 43 of the 115 text models it shares with OpenRouter as a listed provider. | Per-token, routed by OpenRouter on price and uptime |
Private (126 models): Venice runs the model and keeps no prompt data. Anonymized (256): the request goes to a third-party provider with the user's identity removed. E2EE (12 text SKUs): end-to-end encrypted with TEE attestation.
32 models added in the last 30 days, 3 scheduled removals, 3 beta flags. The frontier closed models arrive as anonymized SKUs within days of launch.
Privacy-minded consumers, builders who need uncensored or no-logs inference, and agents: the API takes OpenAI-style calls and DIEM gives agents a budget that does not need a card.
Venice competes in three markets at once. The gap column is where each rival leaves room for Venice.
| Rival | Market | What they do well | Gap Venice fills |
|---|---|---|---|
| ChatGPT, Claude, Gemini, Grok | Consumer frontier AI | Best first-party models, huge distribution, agents and memory. | Chats are stored and moderated by the vendor. Venice offers the same models anonymized, plus private open models, in one app. |
| Proton Lumo, Duck.ai, Brave Leo | Private consumer AI | Trusted privacy brands, free tiers, bundled with existing products. | Small model menus, little or no image, video or API. Venice covers the whole stack. |
| OpenRouter | API router | One key for 469 models, price and uptime routing, huge developer reach. | Privacy depends on the provider picked. Venice sells there too, so it is both rival and channel. |
| Together, Fireworks, DeepInfra, Groq | Open-weight inference | Low per-token prices, enterprise contracts, fine-tuning. | No consumer app and standard content policies. Venice adds uncensored SKUs and a no-logs promise. |
| Phala / RedPill, Near AI | Confidential AI (TEE) | Hardware attestation for private inference. | Venice resells TEE inference as E2EE SKUs inside its catalog, so buyers get attestation without a new vendor. |
| Bittensor (Chutes), Morpheus | Token-backed compute | Decentralised supply paid in tokens. | DIEM gives a fixed $1 a day of credit at catalog prices, simpler to budget than market-priced compute. |
The same model can be bought four ways. Each has its own price source, so each needs an owner.
| Channel | Price source | Who buys | What operations watches |
|---|---|---|---|
| Direct API | GET /api/v1/models pricing in USD and DIEM | Builders, agents, enterprises | Launch prices, removals, docs matching /models |
| OpenRouter | Venice provider entry in OpenRouter /endpoints | Developers already on OpenRouter | Parity with direct, rank by price, 24h uptime |
| Pro Plus and Max | $68 or $200 a month buys $75 or $225 of API spend, banked for two or three months | Power users who also build | Allowance use, banking expiry, upgrade path |
| DIEM | $1 of credit per staked DIEM per day, at catalog prices | VVV holders, agent operators | DIEM price, supply, credit use versus mint |
The desk reads the first two live and models the fourth. Open the channel parity tab ↗
Read from public data on 9 October 2026 and re-checked independently. Each is a routine weekly check, worded as something to confirm; the team may already track all of them.
| Observation | Detail | Suggested owner action |
|---|---|---|
| Venice's own sources agree | The docs pricing page and /models match on all 126 text prices and every privacy label, and the docs trait list matches /models/traits. Retiring models drop off the pricing table as the deprecation policy says. Five audio and decision ids have no docs price row. | Keep it as a launch check |
| Channel prices agree | All 35 models Venice serves both directly and on OpenRouter match on list price. Seven run a temporary OpenRouter discount (5% to 30%), shown in the endpoint's discount field, e.g. GLM 5.3 at $1.40 / $4.40 against a $1.75 / $5.50 list. | Log each promotion's start and end in the weekly brief |
| Compatibility aliases predate the frontier SKUs | /models/compatibility_mapping resolves gpt-4o and gpt-4o-2024-11-20 to llama-3.3-70b, while the catalog now carries openai-gpt-4o-2024-11-20 as its own model. A team migrating from OpenAI that keeps the old name may expect GPT-4o. | Review the alias table at each launch |
| Uptime is scored publicly | OpenRouter publishes each provider's 24h uptime, excluding rate limits and caller errors. On 10 models Venice sat more than 2 points below the provider median on the day checked, e.g. Kimi K2.6 at 89.9% against a 99.7% median. | Add to the weekly provider review |
| Removals inside three weeks | qwen3-5-9b and deepseek-v3.2 on 21 October, claude-sonnet-4-5 on 30 October, announced on the deprecations page. autoRemap is off on all three. | Track each old id to zero before the date |
| Two pages, two welcome credits | The docs pricing page gives new Pro subscribers a one-time $10 credit; the plan table on venice.ai/pricing shows 500 welcome credits, which is $5 at 100 credits to the dollar. | Pick one and update the other |
| Same price, different suffix | Three "Pro" SKUs carry the same price as their base model (GPT 5.6 Luna, Terra, Sol). | Confirm the tier difference is explained |
Every row reproduces in the catalog checks tab ↗, which re-reads the data on each load.
The posting is an open pool with no duty list. These are the business-operations workstreams Venice's model implies, with what I would do in each.
| Workstream | What it covers | My plan |
|---|---|---|
| Catalog and pricing | Model launches, prices in USD and DIEM, removals, descriptions and sources in /models and docs. | A launch and removal runbook (§06), one price sheet that feeds /models, docs and OpenRouter, and the weekly catalog check. |
| Providers and compute vendors | GPU and inference partners, TEE partners, proxied frontier vendors, their uptime and cost. | A provider scorecard: price, uptime, incidents, contract dates. Reviewed weekly with engineering. |
| Channels and partners | OpenRouter, agent frameworks, app integrations, affiliate and referral programmes. | Track listing coverage, parity and rank per channel; one owner and one review per partner. |
| Token operations | VVV staking, DIEM mint rate and supply, buyback and burn reporting, holder communications. | A DIEM and VVV dashboard in the weekly brief; checklists for each emission or mint-rate change. |
| Billing, finance ops and data | Subscriptions, credits, crypto payments, refunds, revenue and usage reporting. | Clean definitions for revenue by line and channel; reconciliations that run themselves. |
| Internal tooling | The scripts and dashboards that keep the above from being manual. | Build them: the desk shows the standard, from live data to a written brief. |
Weekly and written, so the team spends meetings deciding things.
| When | Output | Inputs |
|---|---|---|
| Monday | Weekly brief: catalog changes, removals due, channel parity, provider uptime, DIEM and VVV, open actions with owners | /models, OpenRouter /endpoints, Base token data, support tags |
| Wednesday | Provider review: anything below target uptime or above target cost, with the fix and the owner | Provider scorecard, incident log |
| Each launch | Launch checklist signed off: price, privacy tier, description, source, docs, OpenRouter listing, announcement | Model card, provider terms, price sheet |
| Each removal | Notice under the published 7 to 15 day policy, email to recent callers, successor named, usage of the old id tracked to zero | Usage by model id, customer list |
| Month end | Revenue and usage by line and channel; DIEM credit use versus supply | Billing, chain data |
The ops brief tab ↗ writes the Monday brief from live data today.
Model, provider, privacy tier (private, anonymized or E2EE), licence and terms. Decide the SKU id and display name so it cannot collide with an existing one.
Cost per 1M tokens from the provider, target margin, competitor and OpenRouter prices, then one price in USD and DIEM. Record the reason.
Same price and description in /models, docs and the OpenRouter listing on the same day. Description and modelSource filled.
First two weeks: usage, errors, uptime against the OpenRouter median, support tags. Beta flag removed once stable.
Date set under the published 7 to 15 day policy, longer for models with enterprise usage. Successor named, autoRemap decided on purpose, email to everyone who called the id in the last 15 days, warning in API responses.
Usage of the old id at or near zero, docs and listings cleaned, final note in the weekly brief.
DIEM ties the token to the product. Its price, supply and credit use move together, so they sit in the same read.
| Number | Today | Why it matters |
|---|---|---|
| Collateral per new DIEM | ≈ 432 sVVV (≈ $10,000) | Read from the staking contract's mint curve: about six times the DIEM market price, locked and recoverable. Buying DIEM on market is the cheaper route for a pure compute buyer. |
| Buybacks | ≈ $900,000 in 30 days | DefiLlama holders revenue: VVV bought back and burned, incl. the $414K discretionary burn on 7 October. |
| DIEM price | ≈ $1,650 | Prices a perpetual $1 a day of compute. 365 / price gives the implied yearly credit yield, ~22% today. |
| DIEM supply and staked | ≈ 36,900 minted, ≈ 28,300 staked | Staked DIEM draws $1 a day each: about $28,300 a day, or $10.3M a year, of inference served as credit. The mint rate rises as supply nears the 40,000 target. |
| VVV price and staking share | ≈ $22.8 | Staked VVV is the input for minting; the mint rate rises as DIEM supply grows. |
| Credit use versus allowance | internal | How much of the daily staked allowance is actually used, and on which models. Unused credit costs nothing; used credit on proxied frontier models costs the most. |
Worked example in the desk: 2M input and 400k output tokens a day on GLM 5.2 cost $4.56 a day at list price, covered by 5 DIEM. Open the DIEM planner ↗ · Token ops, live from Base ↗
| Window | Focus | Done when |
|---|---|---|
| Days 1 to 30 | Learn the cadence and the numbers; take one recurring report off someone's plate; write the launch and removal runbook. | One report runs without me; runbook used on a real launch. |
| Days 31 to 60 | One price source of truth across direct, OpenRouter and DIEM; provider scorecard live; weekly brief adopted. | No unexplained channel price gaps; provider review held weekly. |
| Days 61 to 90 | Automate the manual checks; pick up the operating gap the team names first; month-end revenue and usage pack. | Checks run on a schedule; month-end pack delivered on time. |
Built from the public job posting (2026), Venice's public API, pricing page, docs and status page, OpenRouter's public model and endpoint APIs, DefiLlama, and Base contract reads, read on 9 October 2026. The demo is my own tool, built for this application. Counts change daily; the demo re-reads them on each load.
Venice: models API · pricing · docs · blog · Series A · deprecations · alias table
OpenRouter: models API · endpoints example · FAQ (no inference markup)
Status and buybacks: veniceai-status.com · DefiLlama · VVV and DIEM docs
Tokens: DIEM on Base · VVV on Base
Independent homework for the Venice Business Operations open application · 2026 · edwardtay.com