Models & providers

How to specify a model

Always use the creator/model-slug format:

model: "openai/gpt-5.4"
model: "anthropic/claude-opus-4.6"
model: "google/gemini-3.1-pro-preview"
model: "meta/llama-4-405b"
model: "bfl/flux-2-flex"            // image
model: "google/veo-3.1-generate-001" // video

A model id maps 1:1 to its (creator, slug) pair. The full live catalog is at /models — filter by modality, search by capability. With the AI SDK, pass the same id to streamText and swap models without touching the rest of the call.

Modalities

Every model has one or more modalities. Filter or branch on them:

Catalog freshness

The catalog is synced nightly from the routing layer. New models appear within 24 hours of upstream availability. Pricing is the live rate — the number you see on /models is the number you pay.

const res = await fetch("https://synapse.garden/api/v1/models", {
  headers: { Authorization: `Bearer ${process.env.MG_KEY}` },
})
const { data: models } = await res.json()

// Filter by type
const text = models.filter((m) => m.type === "language")
const image = models.filter((m) => m.type === "image")
const video = models.filter((m) => m.type === "video")

const res = await fetch(
  "https://synapse.garden/api/v1/models/openai/gpt-5.4/endpoints",
  { headers: { Authorization: `Bearer ${process.env.MG_KEY}` } },
)
const { data } = await res.json()

console.log(data.architecture.input_modalities)  // ["text", "image"]
console.log(data.endpoints[0].pricing.prompt)    // "0.0000025"

Featured providers

Try a model

Pricing math

Prices on the catalog page are the list price you pay. They include our flat margin baked in — no separate "markup" line on your bill.

For the full math (passthrough + flat 10% DX premium), see /legal/pricing-disclosure.

Choosing a model

Quick heuristics:

• Frontier reasoning — openai/gpt-5.5-pro, anthropic/claude-opus-4.6 with extended thinking
• Production workhorse — openai/gpt-5.4, anthropic/claude-sonnet-4.6
• High-volume cheap — openai/gpt-5.4-mini, google/gemini-2.5-flash
• Edge / latency-sensitive — openai/gpt-5.4-nano, cerebras/*
• Long context — google/gemini-3.1-pro-preview (1M+), anthropic/claude-opus-4.6 (1M variant)
• Image generation — google/gemini-3-pro-image, bfl/flux-2-flex
• Video generation — google/veo-3.1-generate-001, klingai/kling-v2.6-i2v

When in doubt, run the playground above with the same prompt across two or three models and compare.