AionLabs: Aion 3.5

Aion 3.5 is a text model from AionLabs, listed on api14 under the model id aion-labs/aion-3.5. AionLabs describes it as a multi-model roleplaying and storytelling system built on the GLM family of models. Rather than a single monolithic generator, it uses a collaborative generation process in which multiple specialized models each contribute to the output. That design choice shapes what the model is good at, and it is worth keeping in mind when you decide where to place it in a pipeline.

On the input side, Aion 3.5 accepts text only and returns text only, so it is not a fit for audio, image or video work. Its context window is 262,144 tokens, which is large enough to hold long conversations, extended narrative arcs or sizeable reference documents in a single request. The maximum output is 32,768 tokens, so responses can run long when you need them to. The model supports reasoning output, and you can control that behaviour through the include_reasoning, reasoning and reasoning_effort parameters.

Those capabilities point to a fairly specific set of uses. Roleplay and interactive storytelling are the obvious ones, since the model was built for them and the long context lets a session carry a lot of prior turns without trimming. The same properties suit character-driven chat, long-form fiction drafting and collaborative writing tools where a user and the model build a story together. The 32,768 token output ceiling helps when you want a full scene or chapter section in one response rather than stitching together many short completions. Structured output via response_format and function calling via tools and tool_choice also make it usable in applications that need the model to return data in a defined shape or call into your own code.

You call it through api14's OpenAI-compatible API using the model id aion-labs/aion-3.5, so existing OpenAI SDKs and clients work with a change of base URL and model name. The supported request parameters are include_reasoning, max_tokens, reasoning, reasoning_effort, response_format, temperature, tool_choice, tools and top_p. Use temperature and top_p for sampling control, max_tokens to cap response length, and the reasoning parameters when you want to see or steer the model's reasoning. The tools and tool_choice pair covers function calling, and response_format covers structured outputs.

Since the model is text in and text out, tasks that depend on other modalities need a different model. The same applies if you want a general purpose assistant rather than one tuned for roleplay and storytelling, though the tool and structured output support means it can still serve as the generation step inside a larger application.