Models · OpenAI

GPT-OSS 20B, paused on Evanesca.

GPT-OSS 20B is an open-weight model by OpenAI that works with text, released under the Apache-2.0 licence. Paused since 20 Oct 2026: OVHcloud, the only European operator that runs it for Evanesca, is not serving it at the moment. Meanwhile, pick GPT-OSS 120B.

Checked on 20 Oct 2026

GPT-OSS 20B

by OpenAI · Apache-2.0

Gravelines, France Checks 6/6

Input / M: €0.0672
Output / M: €0.252
Quick question: €0.0001

Prices per million tokens, VAT included. Quick question: 300 tokens in, 400 out.

The model at a glance

Maker: OpenAI
Licence: Apache-2.0
Size: 21B (3.6B active) parameters
Context: Not verified yet
Type: Text
Tool calling: Yes measured by a real call
Reasoning: Yes — it thinks before answering those tokens are billed as output
Run by: OVHcloud OVH SAS, France — no US parent company Paused since 20 Oct 2026
Where it runs: Gravelines, France
Checks: 6/6 All six: Licence, Provenance, Security, GDPR, Benchmarks and Live test
Price: €0.0672 per million input tokens · €0.252 per million output tokens VAT included; reasoning tokens count as output
Open weights: Yes published by the maker: an operator can run the model
Model card: huggingface.co/openai/gpt-oss-20b

Where your prompts go

Here is the route of a message sent to GPT-OSS 20B, end to end. No call to the model leaves Europe: not as a backup, not when servers are busy.

From your browser to EvanescaEvanesca passes your message straight on to OVHcloud. It stores neither the prompt nor the answer: both stay in memory only while the answer is written.
OVHcloud, GravelinesWhen it is served: OVHcloud runs GPT-OSS 20B in Gravelines, France. It is a French company with no US parent: out of the CLOUD Act's direct reach. OVHcloud states “zero data retention: we keep only the data required for billing purposes” (OVHcloud AI Endpoints).
Back to you, with its passportThe answer comes back with its passport: model, operator, country, and what was kept. Your history lives in your browser; about this message, Evanesca keeps only billing data — tokens and cost.

What it costs

€0.0672 per million input tokens and €0.252 per million output tokens, VAT included. Three typical exchanges on GPT-OSS 20B:

COST OF AN EXCHANGE · VAT INCLUDED
Example	Tokens in	Tokens out	Cost
A quick question	300	400	€0.0001
A long document to summarise	15,000	800	€0.0012
A message late in a long conversation	40,000	2,000	€0.0032

€2 of credit covers about 16,000 quick questions on this model. Reasoning tokens count as output, so a model that thinks first can cost more than these examples. The exact cost of each answer shows under it. All prices.

Licence

OpenAI publishes GPT-OSS 20B under the Apache-2.0 licence. Apache-2.0 is a permissive open-source licence: it allows commercial use.

The official model card is the reference.

Similar models

Same maker first, then the same type.

Paris, France

GPT-OSS 120B

OpenAI · Text · checks 5/6

€0.252 · €1.008 / M tokens

Germany (AKI.IO)

GLM-5.3

Zhipu / Z.ai · Text · checks 6/6

€3.024 · €9.24 / M tokens

Paris, France

DeepSeek V4 Flash

DeepSeek · Text · checks 6/6

€0.672 · €1.344 / M tokens

Paris, France

GLM-5.2

Zhipu / Z.ai · Text · checks 6/6

€3.024 · €9.24 / M tokens

Paris, France

Qwen3 235B

Qwen · Text · checks 5/6

€1.26 · €3.78 / M tokens

Paris, France

Llama 3.3 70B

Meta · Text · checks 5/6

€1.512 · €1.512 / M tokens

Every model, with its operator and where it runs

Frequently asked questions

Can I use GPT-OSS 20B today?

Not at the moment: it has been paused since 20 Oct 2026, because OVHcloud, the only European operator that runs it for Evanesca, is not serving it for now. A call answers 503 and is not billed. Meanwhile, use GPT-OSS 120B.

Where does GPT-OSS 20B run on Evanesca?

When it is served: In Gravelines, France, on OVHcloud's servers. OVHcloud is a French company with no US parent, so your prompts stay out of the CLOUD Act's direct reach. No call to the model leaves Europe: not as a backup, not when servers are busy.

Does Evanesca keep what I send to GPT-OSS 20B?

No. The prompt and the answer stay in memory only while the answer is written, and your history lives in your browser; about your messages, Evanesca keeps only billing data — tokens and cost. OVHcloud states “zero data retention: we keep only the data required for billing purposes” (OVHcloud AI Endpoints).

How much does GPT-OSS 20B cost?

€0.0672 per million input tokens and €0.252 per million output tokens, VAT included. A quick question — 300 tokens in, 400 out — costs about €0.0001. There is no subscription: credit starts at €2.

Can GPT-OSS 20B call tools?

Yes: a real call through OVHcloud returned proper tool calls (function calling).

Which licence is GPT-OSS 20B under?

Apache-2.0. Apache-2.0 is a permissive open-source licence: it allows commercial use.

Which checks has GPT-OSS 20B passed?

6 of 6: Licence, Provenance, Security, GDPR, Benchmarks and Live test. Its passport shows the same stamps; how each check works.

Does OpenAI see what I type?

No. OpenAI published the weights, and OVHcloud runs them in France: your prompts go to OVHcloud's servers, not to OpenAI.