ScaleDown
Small task-specific models for prompt compression and extraction
scaledown.aiNew York, United States
ScaleDown is a small New York startup that builds task-specific language models served over a REST API. Its products cover prompt compression, summarisation, extraction and classification, aimed at cutting token costs for LLM applications.
- Founded
- 2024
- Type
- Private company
- Founders
- Neal Patel and Soham Chatterjee
- AI lab
- Inference provider
- 3 hosted endpoints
Catalogue
All models by ScaleDown
3 models · 1–3
| Select models to compare | Model | Released | Context | Input /M | Output /M | Parameters | Type | I/O | Openness |
|---|---|---|---|---|---|---|---|---|---|
| SummarizeScaleDown | October 13, 2026 | 1M | $0.04 | $0.00 | — | Text | text, image, file → text | Closed | |
| ExtractScaleDown | October 13, 2026 | 1M | $0.04 | $0.00 | — | Text | text, image, file → text | Closed | |
| Classify-1ScaleDown | October 13, 2026 | 1M | $0.04 | $0.00 | — | Decisions | text, image, file → decisions | Closed |