Upstage: Solar Pro 4

Solar Pro 4 is a text generation model from Upstage, released by the provider on 10 August 2026. It sits in the text category and is described by Upstage as a cost-efficient large language model. If you have used other models from Upstage, this one continues that line of work, aimed at developers who need a general purpose text model rather than a specialised one.

The model takes text in and returns text out, so it is a straightforward language model with no image or audio handling. Its context window is 524,288 tokens, which is large enough to hold entire document sets, long transcripts or sizeable codebases in a single request. The maximum output is 131,072 tokens, so responses can be long as well as inputs. Upstage describes it as built for long-horizon tasks and agentic workflows, with strong capabilities in office productivity and document-intensive work. It also supports reasoning, which you can control through the reasoning and reasoning_effort parameters, and it can call tools.

Those capabilities point to a few practical uses. Document-intensive work is the obvious fit: summarising or extracting structured data from long contracts, reports or research papers that would not fit in a smaller context window. Agentic workflows are another, since the model supports tool calling and parallel tool calls, which suits pipelines where the model decides which functions to invoke. Office productivity tasks, such as drafting, rewriting and transforming business documents, follow from the same strengths. The reasoning controls let you dial effort up for problems that need step by step thinking and down for simple transformations where latency matters more.

You call it through api14's OpenAI-compatible API using the model id "upstage/solar-pro4". 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, max_tokens, parallel_tool_calls, presence_penalty, reasoning, reasoning_effort, response_format, structured_outputs, temperature, tool_choice, tools and top_p. Structured outputs and response_format are useful when you need JSON that conforms to a schema, while tools and tool_choice cover function calling.

One thing to keep in mind is that the model is text only. If your application needs to process images, audio or video directly, you will need a different model for that part of the pipeline. The large context window and long maximum output also mean requests can take longer and consume more of your quota than a small model would, so for short, simple prompts a lighter model may be the more practical choice.