4.5B parameters. Runs on one 24 GB GPU or a 16 GB Mac — no datacenter. A drop-in OpenAI-compatible API, so your existing apps and SDKs work unchanged. Private by design: your data never leaves your machine.
Competes with 10–15× larger models in math, code and reasoning — at a fraction of the cost and power.
Point any OpenAI client at NYMPH. Zero code changes. Chat, tools and streaming, all supported.
Run it on your own hardware. No cloud dependency, no per-token surprises, no data exposure.
A multi-phase harness with tools, retrieval and verification — built for real agentic work, not just chat.
Get a key at nymphtech.com/access, point your base URL at NYMPH, and call it like any OpenAI endpoint.
# pip install openai from openai import OpenAI client = OpenAI( base_url="https://api.nymphtech.com/v1", api_key="nymph_live_...", # from nymphtech.com/access ) resp = client.chat.completions.create( model="nymph-4.5b", messages=[{"role": "user", "content": "Refactor this function for readability."}], ) print(resp.choices[0].message.content)
Run NYMPH AI in the cloud through our API today — or on your own machine with the NYMPH S-Quantum card: persistent memory, full privacy, single-digit watts, always on. Same model, same API, your hardware.