Models

Muse Wonder

Multimodal reasoning, coding and document analysis

Reason over text, images, video, audio and PDFs, with a text response and input/output billing.

Capability notes: 1,048,576-token provider context; text output; up to 8,192 output tokens per request

MODEL ID

Gemini 4 Argon

Multimodal reasoning, coding and document analysis

Google’s reasoning model for software engineering, mathematics and long documents. Combine text, images, video, audio and PDFs in the shared multimodal playground.

Capability notes: 1,048,576-token provider context · text output · platform response limit: 8,192 tokens

MODEL ID

Jev 1.13

Support triage and agent routing

Jev's structured decision model. Ask typed questions about shared state and get answers with probabilities.

Capability notes: Choice, Score and Noul outputs

MODEL ID

API alias: jev-latest

Jev 1.14

Visual classification and agent routing

TypeSafe’s latest structured decision model accepts text, JSON and images, returning typed answers with probabilities.

Capability notes: 262,144-token context · up to 4 PNG, JPEG or WebP images · 4 MiB each · 8 MiB total · 16 MP each

MODEL ID

Liquid d1

Structured classification and scoring

A fast decision model from Liquid, available through Vercel AI Gateway.

Capability notes: Choice, Score and Noul outputs

MODEL ID

Laya

Fast routing and policy checks

A System One evaluation model for structured classification, routing, scoring and guardrail checks through Vercel AI Gateway.

Capability notes: 8,192-token context; 2–100 Choice options

MODEL ID

Solar Decide

Structured classification and scoring

A decision model from Upstage for structured classification, scoring, and yes-or-no decisions.

Capability notes: Choice, Score and Noul outputs

MODEL ID

Solar Decide Flash

Fast text and JSON routing and scoring

Upstage’s low-latency Solar Decide variant, built on Solar Mini 4 for structured text and JSON classification, scoring and yes/no decisions.

Capability notes: 524,288-token provider context · platform: 8 questions, 32 KiB text · no image input · 2–26 Choice options

MODEL ID

Classify-1

Text and JSON classification, routing and moderation

ScaleDown’s structured decision model returns choices, ordered scores and yes/no probabilities for routing, moderation, intent detection and domain tagging.

Capability notes: 1,000,000-token provider context · platform: 8 questions, 32 KiB text · text and JSON; no image or PDF input

MODEL ID

Tev1 4B Experimental

Compact routing and policy checks

An experimental decision model with the Jev contract on OpenRouter. Choice supports up to 20 options.

Capability notes: 32,768-token context; 2–20 Choice options

MODEL ID

Kev 4B

Compact-model classification and scoring

A compact 4B decision model for typed answers and probability estimates.

Capability notes: 8,192-token context

MODEL ID

Perplexity Decider V1 27B

Text, JSON and images

Perplexity’s 27B decision model classifies, routes and scores shared text, JSON or images with typed answers and probabilities.

Capability notes: 262,144 tokens · PNG, JPEG, WebP · up to 4 images · 4 MiB each · 8 MiB total · 16 MP each

MODEL ID

Perplexity Decider V1.1 27B

Text, JSON and image decisions

Perplexity's newer 27B decision checkpoint returns typed answers and probabilities for shared text, JSON or image state.

Capability notes: 262,144-token context · up to 4 images · 4 MiB each · 8 MiB total · 16 MP each

MODEL ID

Perplexity Decider 27B Latest

Text, JSON and image decisions

A rolling Perplexity model alias that currently routes to Decider V1.1. Its target and upstream price can change; text, JSON and image inputs are supported.

Capability notes: Currently 262,144-token context · up to 4 images · 4 MiB each · 8 MiB total · 16 MP each

MODEL ID

Mercury Decide

Text routing, classification and policy checks

Inception's structured decision model returns typed choices, scores and yes-or-no judgments with probabilities.

Capability notes: 65,536-token context · text and JSON · no image input

MODEL ID

Mercury Decide (free)

Routing, classification and policy checks

The explicit free OpenRouter route for Inception’s Mercury Decide. It returns choices, scores and yes-or-no judgments with probabilities. Upstream inference is free; this service charges 1 credit per successful call.

Capability notes: 65,536-token provider context; up to 8 questions; text and JSON

MODEL ID

Microsoft-Decision-1

Routing, classification and policy checks

Microsoft’s Qwen3.5-9B-based decision model returns a calibrated probability for fixed answers in one pass. Use it for classification, routing, prioritization and agent guardrails.

Capability notes: 32,768-token provider context; up to 8 questions; text and JSON

MODEL ID

Drex v1.5

Routing, classification and policy checks

Nace.AI’s compact decision model has fewer than 10B parameters and a 131,072-token provider context. Route, classify and check policies with typed answers and probabilities.

Capability notes: 131,072-token provider context; up to 8 questions; text and JSON

MODEL ID

Drex v2.0

Routing, classification and policy checks

Nace.AI’s 35B mixture-of-experts decision model with 3B active parameters. Score typed answers in one pass for routing, classification and policy checks.

Capability notes: 131,072-token context · text and JSON · no image input

MODEL ID

Clef Omni

Multimodal classification and scoring

Cloudflare’s mixture-of-experts decision model is based on Qwen3-Omni-30B-A3B, with 30B total and 3B active parameters. Classify, route and score text, JSON or images with typed answers.

Capability notes: 65,536-token provider context; up to 8 questions; images supported

MODEL ID

GPT-6 Luna Decisions

Text, JSON and image decisions

OpenAI’s GPT-6 Luna Decisions through the shared Jev interface. Classify, route and score text, JSON or images with typed answers and probabilities.

Capability notes: 1,050,000-token provider context · platform: 8 questions, 32 KiB text · up to 4 images

MODEL ID

Clef Flash

Short text, JSON and image decisions

Cloudflare’s compact decision model checks short text, JSON and images. Use it for screenshot classification, rules and yes/no checks.

Capability notes: 2,000 UTF-8 bytes (text state) · PNG, JPEG, WebP · up to 4 images · 4 MiB each · 8 MiB total · 16 MP each

MODEL ID

Clef

Short text, JSON and image decisions

Cloudflare’s 27B decision model classifies and scores short text, JSON and images through the shared decision API.

Capability notes: 2,000 UTF-8 bytes (text state) · PNG, JPEG, WebP · up to 4 images · 4 MiB each · 8 MiB total · 16 MP each

MODEL ID

Span-01

Evaluate conversational behaviors

A specialized Noul behavior model. Use the Span playground to test behavior judgments.

Capability notes: Noul only; conversation input

MODEL ID

Span-01 Lite

Lightweight behavior checks

A lightweight Span behavior model for Noul yes-or-no judgments.

Capability notes: Noul only; conversation input

MODEL ID

Span-01 Lite (free)

Lightweight behavior checks

The explicit free OpenRouter variant for Noul behavior evaluation. Upstream inference is free; this service charges 1 credit per successful request.

Capability notes: Noul only; conversation input

MODEL ID