Qwen: Qwen3.8 Max Prime

Qwen3.8 Max Prime is a text generation model from Qwen, the model family maintained by Alibaba's Qwen team. It is positioned as a higher-throughput variant of Qwen3.8 Max, served as a separate SKU, and it was released by the provider on 2026-09-23. If you already know the Qwen line, the name tells you most of what you need: this is the Max tier, with the Prime suffix marking the throughput-oriented serving option rather than a different base model.

The model accepts text, image, and video as input and returns text only. That combination makes it a multimodal reader rather than a generator of media: you can hand it documents, screenshots, photos, or video alongside written instructions, and it answers in prose or structured text. The context window is 1,000,000 tokens, which is large enough to hold very long inputs without chunking, and the maximum output length is 131,072 tokens, so long-form responses are practical in a single call. Reasoning is part of the surface area here: the supported parameters include reasoning, reasoning_effort, and include_reasoning, so you can control whether reasoning is exposed and how much effort the model spends on it.

Those capabilities point at a fairly specific set of jobs. Long-context work is the obvious fit: reading entire codebases, contract sets, or research corpora in one request, then asking questions across all of it. Because images and video are accepted as input, the same call can cover mixed material, for example a recording plus its transcript plus a written spec, with the model reasoning over all three. Structured extraction is another natural use, since structured_outputs and response_format are supported, which suits pipelines that need JSON or schema-shaped results rather than free text. Tool use is supported too, so the model can sit inside an agent loop that calls functions or external services.

You call it through api14's OpenAI-compatible API using the model id "qwen/qwen3.8-max-prime". Because the interface is OpenAI-compatible, existing clients and SDKs work with a change of base URL and model id. The request parameters it supports are frequency_penalty, include_reasoning, logprobs, max_tokens, presence_penalty, reasoning, reasoning_effort, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, and top_p. That covers sampling controls, determinism via seed, stop sequences, token limits, log probability inspection, structured output, and tool calling, so most standard request shapes are available without workarounds.

One practical caveat: the provider describes this as a higher-throughput variant served at a higher price point than Qwen3.8 Max, so if raw throughput is not your bottleneck, the standard Max SKU may be the more sensible default. There is no information here about quality differences between the two, so treat the choice as a serving trade-off rather than a capability one. Also note the output side is text only, so anything requiring generated images, audio, or video needs a different model.