Decision Models

Featured

Hosted open decision models behind a Jev-compatible API, with an EU-only tier and a lower-cost Peer-to-Peer tier

Decision Models (decisionmodels.io, formerly System1 Models) hosts open decision models behind a Jev-compatible API: send a state and typed questions, get a probability for every allowed answer. Models include s1-fast and s1-vision (Blink v0.3 26B-A4B), s1-pro (based on vLLM-SR d3) and Quyet-1.0-Large, all Apache-2.0. Each question takes up to 8,192 input tokens.

The EU tier runs on contracted EU operators with a DPA. The cheaper Peer-to-Peer tier can use capacity outside the EU on opt-in keys. From $0.019 per 1M input tokens (s1-fast); output is free.

Pricing: Per token usage

Hosting Cloud

Pricing Usage Based, $0.019 per 1M input tokens (s1-fast), output free

HQ 🇩🇪 Germany

Compliance GDPR

Visit website →

Pricing

Decision Models prices by model

Per 1M tokens, read off Decision Models's own pricing page on the date shown.

Model	Input / 1M	Cached / 1M	Output / 1M	Checked	Notes
Quyet-1.0-Large	$0.035	–	$0.00	15 Oct 2026	Peer-to-Peer rate; EU tier $0.038.
s1-fast	$0.019	–	$0.00	15 Oct 2026	Peer-to-Peer rate; EU tier with a DPA $0.021. Runs Blink v0.3 26B-A4B.
s1-pro	$0.08	–	$0.00	15 Oct 2026	Peer-to-Peer rate for new customers; EU tier $0.084. Existing customers pay $0.038 until 12 Nov 2026. Runs vLLM-SR d3 26B.
s1-vision	$0.031	–	$0.00	15 Oct 2026	Text-only Peer-to-Peer rate; a request with an image is billed $0.217 (EU tier $0.032 and $0.228).

What is Decision Models?

Decision Models (decisionmodels.io) hosts decision models behind an API that accepts TypeSafe Jev's request format. It traded as System1 Models until October 2026, and the old API host, api.system1models.ai, keeps working. Instead of generating text, a model answers typed questions about a state you send: pick one option from a list (choice), give the probability of yes (noul) or rate on a scale you define (score), with a probability for every allowed answer. Output tokens are free because nothing is generated.

Decision Models is a brand of productivity-boost.com Betriebs UG (haftungsbeschränkt) & Co. KG in Passau, Germany. The models are open-weight models from other authors, served on its own engine. The company states it is not affiliated with TypeSafe.

Models

s1-fast: Blink v0.3 26B-A4B (NVFP4), text.
s1-vision: the same Blink model, with or without one image.
s1-pro: a 26B model based on vLLM-SR d3 (FP8). The previous version, Perplexity Decider v1.1, stays available as s1-pro-2 until at least 11 November 2026.
quyet-1-large: Quyet-1.0-Large by Chinh Nguyen (FP8), text.
s1-llm-auto-router: a CPU classifier that answers a fixed set of seven routing questions about a prompt.

All four decision models are Apache-2.0. Each question, including its state, is limited to 8,192 input tokens. The model_version field in a response shows which model answered.

Two tiers

The EU tier runs only on contracted operators inside the EU: the API gateway on Hetzner and GPU inference on UpCloud, with TLS 1.3 to the gateway and a WireGuard tunnel to the GPUs. A DPA is accepted at signup. Prompts and answers are not stored, logged or used for training. The ISO/IEC 27001 certificates belong to the data centres, not to Decision Models.

The Peer-to-Peer tier costs less and starts on spare EU capacity. Two separate per-key switches widen it, and neither is on by default. "Allow worldwide processing" lets a key use GPUs rented from independent providers outside the EU when EU capacity is busy. Network processing sends single-question text requests for s1-fast-1 to miners on Bittensor subnet 92 (the REAX network), with answers spot-checked and Decision Models' own GPUs as the fallback; on those keys the network's providers may store content.

Pricing

Per 1M input tokens, output free, as of October 2026:

Model	Peer-to-Peer	EU
s1-fast	$0.019	$0.021
s1-pro, s1-pro-2	$0.080	$0.084
s1-vision, text only	$0.031	$0.032
s1-vision with an image	$0.217	$0.228
quyet-1-large	$0.035	$0.038
s1-llm-auto-router	$0.001	$0.001

Existing customers keep s1-pro at $0.038 (Peer-to-Peer) and $0.040 (EU) until 12 November 2026. EUR rates are published too. Credit is prepaid through Stripe from EUR 1, with a free allowance by invitation. Automatic top-up is available and off by default.

Integration

Requests go to api.decisionmodels.io, documented in an OpenAPI spec. An MCP server and an llms.txt cover agents.

Who it is for

Teams that want decision models with EU-only processing and a DPA. On s1-fast that costs less than half of Jev's $0.042 per 1M input tokens. The trade-offs are the 8,192-token limit per question, models from other authors and a small company behind the service.

Frequently Asked Questions

How much does Decision Models cost?

Pricing is per 1M input tokens with output free. As of October 2026, s1-fast costs $0.019 on the Peer-to-Peer tier and $0.021 on the EU tier. s1-pro costs $0.080 and $0.084; existing customers keep $0.038 and $0.040 until 12 November 2026. Requests with an image on s1-vision cost $0.217 or $0.228 for all their input tokens.

Does Decision Models have a free tier?

There is no public free tier. A free allowance is given by invitation; otherwise credit is prepaid from EUR 1.

Does my data stay in the EU?

On the EU tier, yes: requests run only on contracted operators in the EU (Hetzner for the gateway, UpCloud for GPU inference), and prompts and answers are not stored. The Peer-to-Peer tier also stays on EU capacity unless a key has worldwide processing or network processing switched on.

Is Decision Models compatible with TypeSafe Jev?

It accepts Jev's request format: a state plus typed questions (choice, noul, score). Each question, including its state, is limited to 8,192 input tokens. Decision Models is an independent service, not affiliated with TypeSafe.

Is Decision Models the same as System1 Models?

Yes. The service traded as System1 Models until October 2026. The company and the API are the same, and api.system1models.ai keeps working alongside api.decisionmodels.io.