The models are available via ScaleDown's own API and through partners including OpenRouter, Vercel, LiteLLM, and LangChain. In accuracy tests on the CLINC OOS intent-detection benchmark, the Classify model posted a score of 95%, outperforming an average of 89.6% from three OpenAI models at 1,810x lower cost. For a limited window of 50 hours, new users receive 50M free input tokens, while OpenRouter offers a promotional rate of $0.04 per million input tokens.
Wednesday, October 21, 2026
Story
ScaleDown Launches AI Models 467x Cheaper Than Frontier LLM APIs
AI8 days ago
A new architecture of task-specific AI models from ScaleDown eliminates the traditional generative loop to handle routine data processing at a fraction of current expenses. The company released a family of Constrained Decoding Models (CDMs) for classification, extraction, and summarization that process up to 1M tokens of multimodal text and image input in a single pass. These specialized tools are 200x faster and 467x cheaper than frontier LLM APIs, priced at $0.05 per million input tokens with no charge for output.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.