SynapseX API models and per-token pricing

This page is the published price list for the SynapseX Platform API. Every model below is served through the same OpenAI-compatible gateway at https://platform.synapsex.ai/api/v1, billed per token against your prepaid credit balance. There is no subscription on the API platform.

Use it to pick the cheapest model that passes your evaluation, and to estimate a workload before you run it.

How much does the SynapseX API cost per token?

Prices are US dollars per one million tokens. Input and output tokens are billed separately, at different rates. Context and max output are in tokens.

Model	Id	Tier	Context	Max output	Input / 1M	Output / 1M
SynapseX Atlas	`synapsex-atlas-7b-v1`	Standard	128,000	16,384	$0.10	$0.40
SynapseX Forge	`synapsex-forge-code-14b-v1`	Standard	256,000	8,192	$0.40	$1.20
SynapseX Axiom	`synapsex-axiom-science-32b-v1`	Standard	32,768	16,384	$0.10	$0.40
SynapseX Sombrero	`synapsex-sombrero-v1`	Standard	32,768	4,096	$0.15	$0.45
SynapseX Centaurus	`synapsex-centaurus-v1`	Standard	131,072	8,192	$0.28	$1.10
SynapseX Quanta	`synapsex-quanta-qcos-32b-v1`	Advanced	131,072	16,384	$0.35	$0.50
SynapseX Vault	`synapsex-vault-enterprise-v1`	Advanced	131,072	16,384	$0.35	$1.40
SynapseX Nexus	`synapsex-nexus-v2`	Premium	131,072	8,192	$0.10	$0.40
SynapseX Oracle	`synapsex-oracle-v1`	Premium	200,000	32,768	$0.75	$3.00
SynapseX Andromeda	`synapsex-andromeda-v1`	Premium	131,072	8,192	$2.75	$8.00
SynapseX Quasar	`synapsex-quasar-v1`	Premium	200,000	32,768	$3.50	$14.00

This table is the price list, not a permission list. An id can be priced and published here, appear in GET /api/v1/models, and still be refused with HTTP 403 when your key sends it to /chat/completions. The 403 body names the model ids your key can actually call — read it instead of working through this table by trial and error. See Models on the API.

Which model should I pick?

Model	Size	What it is for
Atlas	7B	Compact, fast model for general tasks and high volume. Best for everyday automation, classification and quick responses.
Forge	14B	Specialised in code generation, review and DevOps automation. Optimised for software engineering workflows.
Axiom	32B	Scientific reasoning and technical analysis. Ideal for physics, mathematics, genomics and engineering tasks.
Sombrero	15B	Agile multilingual model for medium-context tasks. Compact and fast for high-throughput API workloads.
Centaurus	14B	Powerful reasoning and analysis for STEM, maths and research. Strong on structured data and technical explanation.
Quanta	32B	Focused on quantum computing and QCOS. Optimised for quantum circuit synthesis, hybrid simulations and QCOS adapters.
Vault	32B	Enterprise long-context model for compliance-sensitive workloads. Handles 131K context for document analysis and data extraction.
Nexus	671B MoE	Fast multi-agent orchestrator for code, chat and tool execution. Optimised for agentic pipelines.
Oracle	671B MoE	Advanced reasoning for complex multi-step tasks, mathematics, logic and deep analysis. Top-tier accuracy.
Andromeda	123B	Flagship multilingual model for research, documents and global communication. Best-in-class for long-form generation.
Quasar	200B	Top-tier flagship model for advanced research, complex agentic tasks and frontier reasoning.

If you have no reason to prefer another, start with synapsex-forge-code-14b-v1 — it is the id most likely to be callable on a new key. The full reasoning is on Models on the API.

How does pay-as-you-go credit billing work?

1 credit = US$1.00. Credits are consumed in proportion to token usage at the rates above; input and output are metered separately. There are no monthly minimums and no subscription to buy on the API platform.

Billing is prepaid and fails closed. A workspace with a balance at or below zero gets an HTTP 402 before any model runs, so you cannot overspend:


{"error":{"message":"Insufficient credits","type":"insufficient_quota"}}

Buy credits at /dashboard/credits in the console before your first call — a valid key on its own is not enough. Pack sizes and the full billing model are on Credits and spending.

Estimating a workload

Token cost is linear, so an estimate is arithmetic rather than guesswork. For a request that sends 3,000 input tokens and generates 500 output tokens on Forge:


input:  3,000 / 1,000,000 × $0.40 = $0.0012
output:   500 / 1,000,000 × $1.20 = $0.0006
total per request                 = $0.0018

At 100,000 requests that is $180. Two levers move that number: send fewer input tokens (trim the system prompt and the history you replay), and cap generation with max_tokens.

Are these the same prices as SynapseX Chat?

No. SynapseX Chat at chat.synapsex.ai is a separate product with a separate account, a separate wallet and a different credit unit — chat credits and API platform credits are not interchangeable, and the Chat model catalog is a different, shorter list with synapsex:-prefixed ids. The Chat price table is at Chat models.

Volume pricing

Volume discounts are available for high-usage accounts. Contact api@synapsex.ai with your expected monthly token volume.

Models on the API — read the catalog at runtime with GET /models.
Chat completions and streaming — the endpoint these prices meter.
Credits and spending — packs, balance and the 402 contract.
Cookbooks and recipes — routing cheap traffic to cheap models.