One-Stop Solution for AI Models — from the Requesty blog.

Requesty's August 2026 changelog covers a dense set of platform upgrades for its unified LLM router. On August 19, the team improved Vertex AI auto caching so that cache creation is skipped when the cacheable prefix contains file attachments with non-gs:// URIs, while external URLs, inline/base64 files, and gs:// URIs Additional August 2026 releases include native video input over the Chat Completions API (supporting public video URLs and base64 data, with YouTube links translated to Gemini's native format), discounted-model visibility in the Model Library and a discount percentage field on the /models endpoint to surface DeepSeek o

PitchBook's company profile shows Requesty as a privately held, venture-backed London-based SaaS developer of a large-language-model gateway, founded in 2023 with roughly five employees. The company closed a $3.58M seed round on 26 September 2025, backed by four investors, and is described as generating revenue while s The profile frames Requesty within the broader AI infrastructure and enterprise SaaS venture landscape, pointing to reports such as the August 10, 2026 AI funding analysis showing $407B raised across megadeals and the August 3, 2026 Enterprise SaaS report on VC funding rebounding. Requesty's seed-round cap-table excerp

The Requesty model library lists GPT-5.5 on AWS Bedrock (us-east-1) as a managed endpoint, priced at $4.95 per 1M input and $29.70 per 1M output, reflecting a flat 10% discount versus the OpenAI list rates of $5.50 and $33.00. The endpoint exposes a 1M+ token context (922K input, 128K output) with vision, reasoning, to Workload cost examples on the page illustrate $0.79 for 100K input + 10K output, $7.92 for 1M input + 100K output, and $79.20 for 10M input + 1M output at those rates before caching savings, with cache reads reducing repeated-context costs further. The endpoint is invoked through Requesty's single OpenAI-compatible bas