Venice added GPT-6.1 Terra on release day, with anonymous access: one dollar per million input tokens and five per million output, a 1.05 million context window. In the same pass came Kling 4.0, Kuaishou’s video model: 30 seconds in a single pass, up to 15 reference assets, 10 keyframes, stereo audio with lip-sync. Two announcements in two days which, read together, say more than either does alone.

The price list, actually read

The press release says “one tenth of Astra’s price”. On Venice’s price list Astra costs 12.50 dollars in and 62.50 out per million: the real ratio is roughly one eighth and one twelfth. Enough to justify the choice, not enough to authorise the round figure. The version circulating online said 10 and 50 dollars: it is wrong, and it is worth saying so, because price lists are the only part of an AI product you can verify without trusting anyone.

What buyers are paying for

The difference between Terra on Venice and Terra elsewhere is not the model, it is the surroundings: prompts not stored, no profile, execution in an enclave. For anyone grinding through invoices and documents that is a stronger argument than a benchmark. For anyone who writes it is less romantic but true: anonymity here is a product feature, not a philosophical position.

The flip side of the token

On the economics, DefiLlama’s numbers help keep things grounded. Over the last 24 hours the protocol burned 24,251 dollars of VVV; cumulative fees stand at 3.89 million. TVL is zero by methodology: DefiLlama keeps staking as a separate series, 869.3 million, 71.47% of a market cap of 1.22 billion, with VVV around 25 dollars. Almost three quarters of the token’s value sits still in a staking contract, not in a market.

It has to be said plainly: the burn is a discretionary choice by the company, not a right held by VVV holders. The weekly report from the community’s stats account speaks of credit purchases at a record 2.03 million and signups flat at +1.4%: a company that sells well to the people already using it and struggles to win over the ones who don’t.

Venice has understood that in 2026 price is the most visible feature, and it is using that well. The open question is a different one: anyone can buy the model, nobody can buy the anonymity. The bet is that the second counts for more than the first, and that someone paying one dollar will learn to pay two to stay off the radar.

Verified sources: Venice, GPT-6.1 Terra model page (27 October 2026: $1/$5 per million, 1,050K context, released 26 October 2026); Venice, GPT-6 Astra page ($12.50/$62.50 per million); DefiLlama, Venice Protocol page (27 October 2026: TVL $0, staking 869.3 m, 71.47% of mcap, 24h fees $24,251, cumulative $3.89 m); CoinGecko, VVV page ($25.13, market cap $1.223 bn); OpenRouter and NanoGPT for the Terra price list; Kling documentation for the Kling 4.0 specifications. Discrepancies against the report: Venice’s TVL is not $1.72 bn (that figure is staking plus market cap, and DefiLlama reports TVL 0), and Astra’s price is not $10/50 but $12.50/$62.50.