Basho · 芭蕉
Japanese nuance, in every response.
A model project for Japanese writing, dialogue, and tool use, exploring nuance, restraint, and the distance between speakers.
In development · model inference not released
Explore the model (Japanese) →
AWAI NETWORK / DISTRIBUTED AI
Basho explores nuance in language. Hokusai brings atmosphere to sound and moving images. AWAI develops both from Japanese expression, supported by the Murakumo inference platform.

MODELS BY AWAI.NETWORK
Basho explores nuance in language. Hokusai brings atmosphere to sound and moving images. Both are AWAI development projects rooted in Japanese expression.
Japanese nuance, in every response.
A model project for Japanese writing, dialogue, and tool use, exploring nuance, restraint, and the distance between speakers.
In development · model inference not released
Explore the model (Japanese) →
Atmosphere in sound and motion.
An audiovisual model project connecting Japanese creative direction with composition, timing, sound, and atmosphere.
In development · model inference not released
Explore the model (Japanese) →GIEMON · 儀右衛門 / PHYSICAL AI RESEARCH
Inspired by the inventions of Hisashige Tanaka, known as Karakuri Giemon. A Cosmos-based model research project exploring how things work and move.
Giemon Spatial 2B v0.1 · Experimental
Explore GiemonBuilt on NVIDIA Cosmos
01 / Infrastructure
Our distributed inference service brings together fleet capacity and an API for developers. Model availability and limits vary by endpoint.
02 / Research
Japanese-language model development for text and audiovisual generation. These projects are in development; the names do not imply released model checkpoints.
03 / Partnership
Discuss an inference integration, model evaluation, or commercial arrangement directly with the team operating the service.
Murakumo runs primarily on our distributed fleet. Available models, capacity, and response times vary by endpoint and workload. Choose the service conditions that fit your application.
Check the selected endpoint or checkout for its current price and limits. Partner billing, capacity, and service commitments are agreed separately. “Cheap” is our design direction, not a promise to beat every provider for every workload.
Requests pass through our gateway and the selected inference backend. Operational and billing records are distinct from prompt and output content. We do not claim service-wide zero data retention. Contact us for retention and deletion requirements before sending data that needs specific contractual controls.
This is a registered-office and mailing address, not a data-centre location.
For partnerships, billing, support, or data requests, contact our team. Include the service and request ID where available; do not email passwords, payment-card details, or private prompts.