Decision model APIs and prices
35 decision models from 24 makers; 29 with a hosted price from 21 hosts, checked 14 to 25 October 2026.
A decision model reads a state and returns a typed answer (yes/no, a choice from a list or a score) with a probability for each option, instead of generating text. Most charge for input tokens only. For how they work and when to use one, read decision models explained.
Models
Sorted by the lowest verified input price per 1M tokens. Open models nobody hosts yet come last.
| Model | Weights | Lowest input / 1M | Hosts | Context | Images |
|---|---|---|---|---|---|
Span-02
Respan
|
Open, Apache-2.0 | $0.01 Respan Span | 1 | β | Yes |
pplx-decider flash v1.2
Perplexity
|
Open, Apache-2.0 | $0.01 Perplexity pplx-decider | 1 | β | Yes |
s1-fast
Decision Models (System1)
|
Open, Apache-2.0 | $0.019 Decision Models | 1 | 8K tokens | β |
Laya
NandhaKishorM
|
Open, Apache-2.0 | $0.02 Runware | 2 | β | β |
Liquid d1
Liquid AI
|
Open, LFM Open License v1.0 | $0.02 Liquid AI d1 | 1 | 32K tokens | Yes |
Mercury Decide
Inception
|
Closed | $0.02 Inception Mercury Decide | 1 | β | Yes |
|
Span-01
Respan
|
Closed | $0.02 Respan Span | 1 | 4K tokens | β |
|
pplx-decider v1.2 27B
Perplexity
|
Open, Apache-2.0 | $0.02 Perplexity pplx-decider | 1 | 256K tokens | Yes |
GLiNER 2.5 Decide
Fastino
|
Open, Apache-2.0 | $0.03 Fastino GLiDE | 1 | β | β |
|
s1-vision
Decision Models (System1)
|
Open, Apache-2.0 | $0.031 Decision Models | 1 | 8K tokens | Yes |
|
C
Quyet-1.0-Large
Chinh Nguyen
|
Open, Apache-2.0 | $0.035 Decision Models | 1 | β | β |
Clef-flash
Cloudflare
|
Open, Apache-2.0 | $0.038 Cloudflare Workers AI | 2 | 24K tokens | Yes |
Celeris-1 Decision
Celeris
|
Closed | $0.04 Celeris-1 Decision | 1 | β | Yes |
|
GLiDE
Fastino
|
Closed | $0.04 Fastino GLiDE | 1 | 40K tokens | β |
Nimble-v4-max
Bespoke Labs
|
Open, CC BY-NC 4.0 | $0.04 Bespoke Nimble | 1 | β | β |
|
S
SubQ Snap 1.1
Subquadratic
|
Closed | $0.04 SubQ Snap | 1 | 384K tokens | Yes |
Tev1 4B Experimental
Together AI
|
Open | $0.04 Together AI | 1 | β | β |
Jev 1.14
TypeSafe AI
|
Closed | $0.042 Runware | 2 | 64K tokens | β |
|
Liquid d1-ultra
Liquid AI
|
Closed | $0.042 Liquid AI d1 | 1 | β | Yes |
Microsoft-Decision-1
Microsoft
|
Closed | $0.042 Microsoft-Decision-1 | 1 | 32K tokens | β |
Strom
Uprelic
|
Closed | $0.042 Strom | 1 | 32K tokens | Yes |
Classify-1
ScaleDown
|
Closed | $0.05 ScaleDown Classify-1 | 1 | 1M tokens | Yes |
Drex v1.5
Nace.AI
|
Open, Nace.AI Open RAIL-M | $0.05 Nace.AI Drex | 1 | 128K tokens | β |
|
Drex v2.0
Nace.AI
|
Open, Nace.AI Open RAIL-M | $0.06 Nace.AI Drex | 1 | 128K tokens | β |
|
s1-pro
Decision Models (System1)
|
Open, Apache-2.0 | $0.08 Decision Models | 1 | 8K tokens | β |
| Closed | $0.10 OpenAI Decisions API | 1 | β | Yes | |
Solar Decide
Upstage
|
Closed | $0.10 Upstage Solar Decide | 1 | 512K tokens | β |
|
Clef-Omni
Cloudflare
|
Open, Apache-2.0 | $0.15 Cloudflare Workers AI | 1 | 64K tokens | Yes |
|
Clef
Cloudflare
|
Open, Apache-2.0 | $0.20 Heabsy | 2 | 64K tokens | Yes |
|
Bespoke-Nimble-9B
Bespoke Labs
|
Open, Apache-2.0 | Self-host | β | β | β |
|
J
Kev
Jared Palmer
|
Open, Apache-2.0 | Self-host | β | 64K tokens | β |
|
M
Mapika Decider
Mapika
|
Open, Apache-2.0 | Self-host | β | 32K tokens | Yes |
Strands Decider
AWS (Strands Agents)
|
Open, Apache-2.0 | Self-host | β | β | Yes |
elephant-1
Base Labs (Baseten)
|
Open, MIT | Self-host | β | β | β |
|
mammoth-1
Base Labs (Baseten)
|
Open, MIT | Self-host | β | β | β |
Hosted in Europe
4 hosts headquartered in Europe, cheapest first; each model's price is in the Hosts table below. Headquartered in Europe says nothing about where requests run: check each host's data processing terms. More European AI providers.
| Host | Models | Input / 1M | HQ |
|---|---|---|---|
| s1-fast, s1-vision, Quyet-1.0-Large, s1-pro | from $0.019 | DE | |
| Strom | from $0.042 | DE | |
| Clef-flash, Laya | from β¬0.042 | SE | |
| Clef | from $0.20 | SK |
Hosts
Every verified hosted price, per 1M tokens, read off the host's own pricing page. The date links to that page.
| Host | Model | Input / 1M | HQ | Checked |
|---|---|---|---|---|
| Clef-flash | β¬0.042 β $0.0477 Preview. Output tokens are tracked but free. | SE Europe | 15 Oct 2026 | |
| Laya | β¬0.042 β $0.0477 Output tokens are tracked but not billed | SE Europe | 15 Oct 2026 | |
| Nimble-v4-max | $0.04 Prepaid credit; new accounts get $1. Output tokens are free. | β | 25 Oct 2026 | |
| Celeris-1 Decision | $0.04 Output free, explanations included. Images count as input tokens. | US | 25 Oct 2026 | |
| Clef | $0.24 Output tokens are free. | US | 15 Oct 2026 | |
| Clef-Omni | $0.15 Output tokens are free. | US | 15 Oct 2026 | |
| Clef-flash | $0.038 Output tokens are free. | US | 15 Oct 2026 | |
| Quyet-1.0-Large | $0.035 Peer-to-Peer rate; EU tier $0.038. | DE Europe | 15 Oct 2026 | |
| s1-fast | $0.019 Peer-to-Peer rate; EU tier with a DPA $0.021. Runs Blink v0.3 26B-A4B. | DE Europe | 15 Oct 2026 | |
| s1-pro | $0.08 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. | DE Europe | 15 Oct 2026 | |
| s1-vision | $0.031 Text-only Peer-to-Peer rate; a request with an image is billed $0.217 (EU tier $0.032 and $0.228). | DE Europe | 15 Oct 2026 | |
| GLiDE | $0.04 Output tokens are free. | US | 25 Oct 2026 | |
| GLiNER 2.5 Decide | $0.03 Output tokens are free. | US | 25 Oct 2026 | |
| Clef | $0.20 Served in the EEA, requests not stored. Output tokens are free; API access on request. | SK Europe | 19 Oct 2026 | |
| Mercury Decide | $0.02 50% promotion, list $0.04. Output not charged. | US | 25 Oct 2026 | |
| Liquid d1 | $0.02 Output $0.00. $2 free credit for new organizations. | US | 25 Oct 2026 | |
| Liquid d1-ultra | $0.042 Output $0.00. $2 free credit for new organizations. | US | 25 Oct 2026 | |
| Microsoft-Decision-1 | $0.042 On Microsoft Foundry. Output free. | US | 15 Oct 2026 | |
| Drex v1.5 | $0.05 Output not billed. $25 signup credit. | US | 15 Oct 2026 | |
| Drex v2.0 | $0.06 Output not billed. $25 signup credit. | US | 15 Oct 2026 | |
| OpenAI Decisions (gpt-6-luna) | $0.10 Generally available. No output charges; cache reads and writes follow the gpt-6-luna caching rates. | US | 25 Oct 2026 | |
| pplx-decider flash v1.2 | $0.01 Output tokens are free. | US | 25 Oct 2026 | |
| pplx-decider v1.2 27B | $0.02 Output tokens are free. Requests naming v1.1 or v1 are billed as v1.2. | US | 25 Oct 2026 | |
| Span-01 | $0.02 span-01-pro; span-01-free is $0. | US | 15 Oct 2026 | |
| Span-02 | $0.01 On the Respan Gateway. Output tokens are free. | US | 25 Oct 2026 | |
| Jev 1.14 | $0.042 Listed as Jev, input tokens only. | US | 15 Oct 2026 | |
| Laya | $0.02 Output tokens are not billed. | US | 14 Oct 2026 | |
| Classify-1 | $0.05 Text input; image input $0.15. Output free. | US | 25 Oct 2026 | |
| Strom | $0.042 Output tokens are free. In beta, access granted manually. | DE Europe | 15 Oct 2026 | |
| SubQ Snap 1.1 | $0.04 Public preview. Output tokens are free; 100M free tokens for new users. | US | 25 Oct 2026 | |
| Tev1 4B Experimental | $0.04 Listed as Tev1 4B Experimental. Output tokens are free. | US | 25 Oct 2026 | |
| Jev 1.14 | $0.042 Output tokens are free. | US | 15 Oct 2026 | |
| Solar Decide | $0.10 Beta. Output tokens are free. | KR | 15 Oct 2026 |
Frequently asked questions
What is a decision model?
A decision model reads a state and typed questions (yes/no, a choice from a list or a score) and returns a probability for each option instead of generating text. All questions in a request are answered in one pass, which is why most hosts charge for input tokens only.
Which decision model API is cheapest?
Span-02 on Respan Span has the lowest verified input price, $0.01 per 1M tokens, checked 14 to 25 October 2026. Output tokens are free on every host listed.
Which decision models have open weights?
Span-02, pplx-decider flash v1.2, s1-fast, Laya, Liquid d1, pplx-decider v1.2 27B, GLiNER 2.5 Decide, and s1-vision and 15 more. Open weights can be downloaded and self-hosted; licence terms differ, so read each model's card.
Which decision model APIs are hosted in Europe?
Hosts headquartered in Europe: Decision Models (DE), Strom (DE), Berget AI (SE), and Heabsy (SK). The lowest verified input price is $0.019 per 1M tokens for s1-fast on Decision Models. Headquarters says nothing about where requests run, so check each host's data processing terms.
How are these prices checked?
Each price is read off the host's own pricing page and dated, checked 14 to 25 October 2026. Prices and sources are open data under CC-BY, as CSV and JSON.
Quality, routers and the raw data
Prices here say nothing about accuracy. For accuracy and calibration scores, see the Artificial Analysis Decision Index or the open Decision Index on Hugging Face. OpenRouter also serves many of these models through one API key; routers pass through the host's price plus a fee, which is why they are not listed as hosts.
The prices are open data under CC-BY: CSV or JSON. For text-generating models, see LLM API pricing by model.
Is your product missing?