Inference Platform

Dedicated endpoints for every agent — in minutes.

The inference backbone behind Synaptix Agent Platform. Open-source LLMs as dedicated, agent-optimized endpoints with guaranteed latency, private networking and per-workflow routing.

Deploy an endpoint See benchmarks See the Agent Platform

< 2 min

To provision a dedicated endpoint

40+

Open & frontier models

99.99%

Enterprise SLA

−50%

Batch inference savings

Inference Platform services

Four ways to run inference for production agents.

From a shared API call to reserved multi-region endpoints — scales with your agent estate.

Real-time inference

Sub-second latency for interactive agents.

Batch inference

Async processing at up to 50% lower cost.

Post-training & fine-tuning

Fine-tune, distill and align open models on your data. LoRA, SFT, DPO and RL.

Dedicated agent endpoints

Reserved capacity in minutes.

Fast deployment

Dedicated agent endpoints — in minutes, not weeks.

Pick a model, region and latency target from the console — the endpoint goes live and agents start routing traffic immediately. Private networking, custom SLAs and full audit from minute one.

Request a demo See latency benchmarks

Sub-2-minute provision

Choose model, region and SLA. Live and routing immediately.

Agent-optimized routing

Per-workflow model selection — reasoning, code, long-context or multimodal.

Private & sovereign

VPC peering, BYOC, air-gapped on-prem. Nothing leaves your perimeter.

Elastic scaling

Autoscale from zero to thousands. Reserved capacity for predictable peaks.

Available models

A catalog that ships with the frontier.

New open releases added within days — same API, no migration.

gpt-oss-120B

General · Reasoning

128K context

Kimi-K2.5

Long context · Agents

2M context

Qwen3-Coder-480B-A35B-Instruct

Code · MoE

256K context

GLM-5

Multilingual · Reasoning

128K context

DeepSeek V3.2

Reasoning · MoE

128K context

MiniMax M2.1

Multimodal · Agents

1M context

Nemotron 3 Super

Enterprise · Reasoning

128K context

Llama 4 405B

General purpose

256K context

Mistral Large 3

European · Tooling

128K context

…and more added every month.

Quickstart

OpenAI-compatible. Drop-in in 3 lines.

Pythonapi.synaptix.ai

from openai import OpenAI

client = OpenAI(
 base_url="https://api.synaptix.ai/v1",
 api_key="sx_live_…",
)

resp = client.chat.completions.create(
 model="gpt-oss-120b",
 messages=[{"role": "user", "content": "Summarize this report."}],
)
print(resp.choices[0].message.content)

cURLapi.synaptix.ai

curl https://api.synaptix.ai/v1/chat/completions \
 -H "Authorization: Bearer $SX_API_KEY" \
 -H "Content-Type: application/json" \
 -d '{
 "model": "deepseek-v3.2",
 "messages": [{"role":"user","content":"Hello"}]
 }'

Pricing