Tencent: Hunyuan A13B Instruct

Tencent Hunyuan A13B Instruct is a text generation model from Tencent, released on 8 July 2025. It is a Mixture-of-Experts (MoE) language model with 13B active parameters and a total parameter count of 80B. Alongside standard instruction following, it supports reasoning through Chain-of-Thought, which makes it useful for tasks where the model benefits from working through intermediate steps before producing a final answer.

The model accepts text as input and returns text as output. It has a context window of 131,072 tokens and a maximum output of 117,964 tokens, so it can handle long documents and produce long responses in a single request. Its knowledge cutoff is 31 March 2025. These characteristics place it in the range of models suitable for substantial single-pass work rather than only short prompts.

Because of the large context window and the long maximum output, typical use cases include summarising or analysing lengthy documents, extracting structured information from long inputs, and generating extended technical or editorial drafts without chunking. Chain-of-Thought reasoning suits multi-step problems such as planning, multi-part question answering, and tasks where a visible or internal reasoning trace improves reliability. Structured outputs and response_format support also make it a reasonable fit for pipelines that need machine-readable results rather than free text.

You call the model through api14's OpenAI-compatible API using the model id "tencent/hunyuan-a13b-instruct". The request parameters it supports are frequency_penalty, include_reasoning, max_tokens, reasoning, response_format, structured_outputs, temperature, top_k, and top_p. If you want the model's reasoning behaviour surfaced or controlled, include_reasoning and reasoning are the relevant parameters; for deterministic formatting, response_format and structured_outputs are available. Because it is a text-only model, inputs must be text and outputs will be text.

Given the text-only modality, another model would fit better if your application needs image, audio or other non-text inputs, since those are outside what Hunyuan A13B Instruct accepts. It supports a broad set of sampling and formatting parameters, so most standard text generation workloads can be configured directly rather than requiring workarounds.