SenseChat

What SenseChat is

SenseChat is a conversational AI product from SenseTime, unveiled on 10 April 2023 as part of the company's SenseNova model family. At launch the chatbot had more than 100 billion parameters ^[1]. Parameters are the numbers a neural network learns during training; roughly speaking, a model with more of them can hold more patterns from its training data, though the count alone does not settle how well a model performs.

Open to organisations first

The first release was not a public consumer product. At launch SenseChat was available only to government and corporate clients ^[1]. Delivering a model to organisations before opening it to everyone is a common route for large models: the use cases are comparatively well defined, and the provider can control who has access and support them directly.

SenseChat V4

In February 2024 SenseTime released SenseChat V4. The version came in three variants with different context lengths — 4K, 32K and 128K — alongside specialised models for data analysis, medicine and vision ^[2]. Context length sets how much text a model can hold in view at one time: a 4K variant suits short exchanges and short documents, while a 128K variant can work through a long document or a long multi-turn conversation without pushing the earlier part out of view.

SenseTime said V4 performed neck-and-neck with GPT-4 and surpassed GPT-3.5 in overall performance ^[2]. The release also introduced a Function Call and Assistants API, which let developers connect the model to outside tools and build assistant-style applications, plus a data-analysis tool called Office Raccoon ^[2]. Function calling is best understood as letting the model ask an external function or service for something it needs, instead of answering only from what it absorbed in training. In the same month, the company said SenseChat had supported more than 500 customers across industries ^[2].

SenseNova V6

With SenseNova V6 in April 2025, SenseChat received a comprehensive update, and a new app went into preview; V6 could also be tried on the SenseChat web platform ^[3]. The chatbot thus moved from being a model offered to business clients toward a product reachable through a browser and a dedicated application.

Where it sits

The sequence of releases describes a fairly legible path. SenseChat began as a general conversational model above the hundred-billion-parameter mark, restricted to government and corporate customers ^[1]. It then widened along three axes: longer context, specialised models for particular kinds of work, and developer-facing interfaces for building on top of it ^[2]. Finally it was refreshed together with the underlying SenseNova system and extended to a web platform and a new app ^[3]. For readers following the field, SenseChat is a case of how a large model becomes a product line — with editions, tools and channels added version by version — rather than a single launch.