SenseTime’s SenseNova Opens a Free, Multimodal API During Public Beta

01 Introduction

Previously, I shared an article about a company called Agnes-ai that provides free models. Today, I discovered another company whose models are in public beta and completely free: SenseTime's SenseNova (SenseNova).

02 Overview

2.1 Introduction

SenseTime's "SenseNova" (SenseNova) large model platform is a force to be reckoned with in the domestic large model camp. From the early SenseChat series to the latest SenseNova 6.8 Flash Lite Preview, SenseTime has taken a "native multimodal + agent" route — not just wrapping a picture interface on top of a text model, but being multimodal at the architectural level, capable of reading documents, viewing images, analyzing tables, and generating deliverables rich in text and graphics.

SenseNova is a Model-as-a-Service platform built by SenseTime on its self-developed large-scale infrastructure (SenseCore) computing base. The platform provides various models including general large models, multimodal large models, and anthropomorphic conversational large models, all served externally via API.

In 2026, SenseTime launched the Token Plan, making all model calls completely free during the public beta period, with 1,500 calls every 5 hours — no prepayment, no trial credits, no expiry tricks. Register and use it directly, grab an API Key and start calling. This article will walk you through the entire access process from start to finish.

Official website: https://www.sensenova.cn/

2.2 Supported Models

Includes multimodal, text-to-image, and more.

03 Free Access

3.1 Registration

Simply go to https://platform.sensenova.cn/docs to register.

3.2 Create an ApiKey

After registration, create your own ApiKey.

04 Connecting to an Agent

Here we choose to connect to OpenCode. Just follow the official documentation.

4.1 Required Parameters

Summarize the main parameters:

• • ApiKey
• • BaseUrl: https://token.sensenova.cn/v1
• • Model parameters refer to section 2.2

4.2 OpenCode Configuration

OpenCode configuration can be done by throwing the documentation at OpenCode and letting it configure itself, or you can manually add a provider to complete the configuration.

Below is the manual approach:

4.3 Testing

sensenova-u1.5-lite and sensenova-u1-fast are text-to-image models and cannot be used directly; we can install custom tools through OpenCode.

We can first ask OpenCode whether it supports text-to-image models:

Then choose your preferred method and install it.

Test the text-to-image effect:

Result:

The effect is quite good.

05 Summary

SenseNova is completely free during the Token Plan public beta, with a generous quota of 1,500 calls every 5 hours — just register and use it.

If you are looking for a free, multimodal domestic large model platform with agent ecosystem support, SenseTime SenseNova is worth trying. Take advantage of the free public beta and get running.