Summarize
Neural Network
Summarize is an abstractive summarization model by ScaleDown for compressing long texts and documents into a coherent summary.
Specifications
- Input
- text, images, documents
- Output
- text
- Context size
- 1 000 000 tokens
- Max answer length
- 4 000 tokens
Prompt cost
(per 1M tokens)
How it works Summarize?
Summarize is an abstractive summarization model by ScaleDown. It does not extract phrases from the source but rewrites the content, resulting in a coherent summary that preserves the original meaning; extractive approaches, by contrast, assemble summaries from quotes and read disjointedly. The model is designed for long texts and documents that need a compact and reliable summary. Through BotHub, the Summarize API is available without a VPN or foreign card: pay with a Russian card in rubles as you go, no subscription, tokens do not expire. Over 250 models are available in one window, and you can switch between them using a single OpenAI-compatible API without rewriting your integration; data is encrypted via AES-GCM and not stored, and businesses have access to contracts, invoices, electronic document management, and an admin panel with limits. It is useful for extracting summaries from reports and research, transcribing meeting and call notes, compiling news digests, preparing annotations for knowledge bases, and compressing voluminous documents before the next pipeline step.Frequently asked questions about Summarize
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
This is a text model from ScaleDown with a one-million token context window. It is useful where you need to compress a large volume of material: summarize a report, extract the essence from correspondence or a contract, or compile an article summary. You can input text, documents, and images.
Up to 4000 tokens per generation — enough for a detailed summary, outline, or structured abstract of several pages. If you need longer output, break the task into parts and ask the model to continue from where it left off.
A separate step-by-step reasoning mode is not noted in our data. In practice, you can get a more thoughtful result via a prompt: ask the model to first highlight key points and the document structure, and then write the final summary.
Support for function calling and strict structured output is not noted in our data. If you need a JSON response, describe the schema directly in the prompt and provide an example — this is usually enough to parse the result on the application side.
Yes, the model accepts images and documents, including native formats — you can send a scan, PDF, or screenshot and ask for a summary of the content. There is no information about audio and video support in our data, so you should not count on them.
The quality of Russian is not separately described in our data, so it is more reliable to test the model with your own material. In BotHub, it's simple: upload a sample document, check the result, and if necessary, switch to another model in the same window without editing your integration.