Evanesca vs Together AI: the same open models, other prices, other defaults

At a glance

Each Together AI cell links to the Together AI page it comes from, read on 6 October 2026; each Evanesca cell links to the page that describes it.

Prices on the models both serve

These open-weight models from Evanesca's catalogue are also on Together AI's serverless price list. The table puts both prices side by side; the Together AI column is copied from its pricing page.

Together AI publishes its prices in US dollars, before tax; Evanesca's are in euros, margin and 20% VAT included. So the verdict compares both before tax: Evanesca's price divided by 1.2, and the dollar converted at the ECB reference rate of 6 October 2026, 1 EUR = 1.1269 USD (ECB). A gap under 2% names no winner.

Together AI is cheaper on both input and output for 6 of the 6, often by a wide margin: on Qwen3 235B, Together AI costs 83% less on both input and output, though its row is an FP8 version tuned for throughput. On Llama 3.3 70B, Together AI costs 27% less on both input and output.

Evanesca's prices follow what its European operators charge — Scaleway in Paris, for every model in this table — plus a margin and VAT. A provider that runs its own GPUs across many customers can price the same weights lower; on price alone, Together AI usually wins.

What is kept, and by whom

Together AI's documentation is plain about its default: “By default, Together stores the prompts you send and the responses models return, and may use them for product improvements.” Stored data is not shared with third parties and can be deleted source.

« Organization admins can turn storage off in privacy settings, which enables zero data retention (ZDR): request content is then not persisted or used for any secondary purpose. »

Two more settings matter. Training on your organisation's data is off by default and opt-in only. “Passthrough” models are on by default: for them, Together AI forwards prompts to an upstream provider whose own policy applies, and turning storage off also turns them off source.

On Evanesca it is the other way round, and there is no switch at all: no prompt and no answer is stored, your history stays in your browser, and only billing data — tokens and cost — is kept. The EU operators state their part: OVHcloud keeps “only the data required for billing purposes” source; Scaleway may keep and read requests that cause errors or look abusive, for up to 14 days source.

Where the model runs, and under which law

On Together AI's serverless API, “the serving region is not selectable” source. Its documentation gives an example: “DeepSeek models are hosted in Together's secure North America data centers” source. On Evanesca, DeepSeek V4 Flash — the same weights — runs with Scaleway, in Paris.

Together AI does offer the EU: it has EU data centres, and dedicated endpoints can be placed there on its Scale and Enterprise tiers, through sales source. That keeps the computation in Europe. It leaves the company where it is: Together Computer, Inc. is a US corporation, and the US CLOUD Act lets US authorities require a provider under US jurisdiction to disclose data in its “possession, custody, or control”, wherever stored source.

Evanesca's models run with OVHcloud in Gravelines or Scaleway in Paris, with Mistral AI in its EU region, with AKI.IO in Germany, or with Infomaniak in Switzerland. OVHcloud, Scaleway, Mistral AI and Infomaniak are European companies with no US parent, out of the CLOUD Act's direct reach; AKI.IO is a German company whose shareholders are not published. Mistral AI keeps its models' inputs and outputs 30 days to monitor abuse source. Evanesca serves no model outside Europe. The sovereignty page gives the details.

Where Together AI is the better choice

• Price: cheaper on most of the open-weight models both serve, sometimes several times over (table above).
• Fine-tuning and custom training, available today; on Evanesca, fine-tuning is planned source.
• Dedicated endpoints, provisioned throughput, batch inference and GPU clusters, for steady high volumes source.
• A much wider catalogue: chat, images, video, audio, embeddings and reranking, from many model families source.
• EU placement for dedicated endpoints on its higher tiers, for teams that need EU residency at scale source.

It is also running today; Evanesca has not opened yet.

Where Evanesca fits better

• No prompt stored, with nothing to configure privacy.
• Every model runs in Europe, by a European operator, on every request sovereignty.
• Prices in euros with VAT included, no subscription: credit from €2 that never expires pricing.
• A chat and an Agent mode for people who do not write code, with a passport under each answer naming the model, operator, country and law Agent mode · passport.
• For companies: a signed Art. 28 DPA, SSO and roles, spending limits business.

Evanesca's API covers chat completions, the model list and embeddings, OpenAI-compatible, with embedding models run in France embeddings.

How to choose

• You need fine-tuning, dedicated capacity or the lowest price per token: Together AI.
• You send personal, client or regulated data and want it processed in Europe by a European operator, with no prompt kept: Evanesca's models.
• You want both: they share the OpenAI API shape, so the same code can call either by changing the base URL and the key.

If you stay on Together AI for sensitive data, switch storage off first: zero data retention is one admin setting away source.

How this page was checked

Together AI's pages were read on 6 October 2026; its prices and settings may have changed since. Evanesca's prices come from its catalogue and every verdict above is computed from them, with the dollar at the ECB reference rate of that day. Spotted an error? Write to us.

• Together AI — pricing
• Together AI — privacy and security (docs)
• Together AI — serverless overview, region selection
• Together AI — EU data centers and dedicated model deployment
• Together AI — privacy policy, updated 17 December 2025
• Together AI — terms of service, updated 19 May 2026

Frequently asked questions