SubQ Snap

The first decision model built for real enterprise work.

SubQ Snap is a multimodal decision model with a 384k-token context window
built for fast, long-context reasoning and agentic decisions in enterprise workflows.

Free 100M tokens for new users
Built by researchers and engineers from
  • Meta
  • Google
  • Oxford
  • Cambridge
  • MIT
  • BYU
  • DoorDash
  • Amazon
  • LlamaIndex
  • TwelveLabs
  • Airbnb
Core Benefits

Built for structured decisions in agentic workflows.

384KTokens of context

Make better decisions with long, multimodal context.

Today's decision models see a sliver of a real codebase. Snap handles up to 384K, enough for the whole repo, data room or agent history, so every call is grounded in all the context it needs.

$0.04Per 1M tokens

Stop paying frontier prices for routine decisions.

Most agent steps are simple calls: pick a tool, open a file, check a result. Snap handles them for $0.04 per 1M tokens and leaves only the hard work to your LLM.

<1sFor 100K tokens

Decisions in seconds, not minutes.

A decision that takes an LLM minutes takes Snap seconds. Your agents move faster without giving up a single point of accuracy.

Benchmarks

The industry leader in short and long context decision-making.

Artificial Analysis Decision Index v2.1

  1. d1.2-ultra75
  2. Drex Max v1.173
  3. SubQ Snap 1.273
  4. GLiDE v1.273
  5. Mammoth 27B71
  6. Decider v1.370
  7. d1.269
  8. Microsoft-Decision-1.168
  9. Elephant 9B66
  10. Jev 1.1466
  11. Gemini Decision 166
Source: Artificial Analysis. Document decisions and extraction, tool calling, knowledge, reasoning and invariance.

Long document decisions

LongBench v2 items up to 256K tokens · InfiniteBench on 3 core tasks
Decision LCR 1.1 Multi-Document
Snap 1.268.4
Snap 1.163.2
Jev 1.1450.5
d1-ultraExceeds context–
Decision LongBench v2
Snap 1.266.3
Snap 1.165.3
Jev 1.1460.5
d1-ultraExceeds context–
Decision InfiniteBench
Snap 1.285.9
Snap 1.182.3
Jev 1.1478.2
d1-ultraExceeds context–

Hugging Face Decision Index 0.3

Public suite · overall index and five areas
AreaSnap 1.2Snap 1.1Jev 1.14d1-ultra
Overall index69.166.460.768.1
Knowledge & Reasoning11 datasets57.855.856.956.6
Language Understanding9 datasets73.069.661.073.5
Retrieval & Classification6 datasets70.967.863.668.6
Tools & Automation5 datasets85.683.974.681.5
Arts & Human Taste6 datasets54.251.239.158.8

Enterprise workflows

16 benchmarks · Subset of HF Decision Index 0.3
BenchmarkSnap 1.2Snap 1.1Jev 1.14d1-ultra
Average75.472.365.373.4
BFCL95.595.495.394.3
API-Bank85.885.686.688.2
ToolRet64.357.259.663.3
When2Call91.691.075.986.1
BANKING7793.592.679.786.5
CLINC150+OOS93.292.990.093.0
PhishNChips64.754.767.561.2
Amazon ESCI54.252.545.552.4
BenchmarkSnap 1.2Snap 1.1Jev 1.14d1-ultra
BRIGHT39.437.740.641.0
HoVer77.672.254.074.3
RAGTruth70.369.448.165.0
ContractNLI75.874.556.872.4
FinEntity91.190.082.191.6
NLI4CT73.871.870.875.3
ACOS53.740.930.950.5
CRUXEval81.977.760.879.1
Use cases

Built with SubQ Snap

Context curation within coding agents
Multimodal recommendation system
Support queue triage
Pricing

Efficiency on Demand.

Free 100M tokens for new users
How it works

What is a long-context decision model?

Decision models are AI models that evaluate supplied context against defined questions and return structured answers. They excel when the deep reasoning or nuanced outputs of a frontier LLM are not required.

DimensionsSubQ Snap 1.2Language model
OutputA choice, score, or labelOpen-ended text
Answer spaceFixed, defined up frontUnbounded
InferenceSecondsMinutes for the same task
Parsing needsNone. Structured output your code uses directlyParse and validate free text
Best atRouting, ranking, classifying, yes/no callsReasoning, writing, open-ended work
Weak atOpen-ended generationFast, repeated, structured calls
Place in agentThe decision points: route, filter, scorePlanning and generation
Enterprise

Interested in
enterprise access?
Contact sales.

We'll use your details to reply to your request. See our privacy policy.
Research

From the lab.

SubQ Snap 1.2: Order-Robust Decisions and Higher Long-Context Accuracy
Product · November 11, 2026
SubQ Snap 1.2: Order-Robust Decisions and Higher Long-Context Accuracy
Read more →
Introducing SubQ Snap 1.1: The Smartest, Enterprise-Ready Decision Model
Product · October 22, 2026
Introducing SubQ Snap 1.1: The Smartest, Enterprise-Ready Decision Model
Read more →
Introducing SubQ Snap: A Long-Context Decision Model for Enterprise Agents
Product · October 12, 2026
Introducing SubQ Snap: A Long-Context Decision Model for Enterprise Agents
Read more →
Introducing SubQ 1.1 Small
Research · June 16, 2026
Introducing SubQ 1.1 Small
Read more →

Building a radically more intelligent world

Subquadratic is a frontier AI research company that believes the architecture layer is the highest-leverage surface in AI. While other major labs focus on incremental improvements to Transformer models, we're pushing foundational change at the architecture level to build models and products that scale efficiently.

Built by researchers and engineers from
  • Meta
  • Google
  • Oxford
  • Cambridge
  • MIT
  • BYU
  • DoorDash
  • Amazon
  • LlamaIndex
  • TwelveLabs
  • Airbnb