Every GPU workload, rupee-billed by the second.
Rent a GPU from the marketplace and run real AI/ML work in your browser — training, image generation, LLMs and more. No setup, no cloud bill.
Available now
Run it today
Live on every GHive host, in a browser notebook — nothing to install.
Fine-tune & train models
LoRA/QLoRA fine-tunes, vision models, and any PyTorch or CUDA job — straight from a browser notebook.
PyTorch · CUDA · JupyterLab · Hugging Face
Bring your own model
Download and run anything — your repo, weights, or workflow — with a real GPU, terminal, and file browser.
JupyterLab · full shell · SSH key
Image Generation (Stable Diffusion, FLUX)
Text-to-image, inpainting and LoRA training in ComfyUI — no local VRAM required.
ComfyUI
LLM Inference
Run open models like Llama, Mistral and Qwen on a real GPU, in your browser.
text-generation-webui
AI Chatbots
Prototype chat assistants on open LLMs in a ready-made web UI — no setup.
text-generation-webui
Coming soon
One-click templates, landing shortly
Curated, hardened images we’re rolling out next — each becomes a one-click launch.
Fine-tuning Models
LoRA/QLoRA and full fine-tunes on 24–80 GB cards without the cloud bill.
Hugging Face PEFT · Axolotl
Video Generation
Generate short clips with open video models on high-VRAM GPUs.
AnimateDiff · open video models
Computer Vision
Train and run detection, segmentation and OCR pipelines.
YOLO · OpenCV · Detectron2
Speech Recognition
Transcribe and translate hours of audio with Whisper-class models.
Whisper · faster-whisper
Reinforcement Learning
Train agents in simulated environments with GPU-accelerated rollouts.
Gymnasium · Stable-Baselines3
Scientific Computing
GPU-accelerated simulation, numerics and large-scale data crunching.
CuPy · JAX · RAPIDS
3D Rendering
Render Blender scenes on fast GPUs, billed only while you compute.
Blender
CUDA Development
Write, build and profile CUDA kernels on real dedicated hardware.
CUDA Toolkit · Nsight
PyTorch / TensorFlow Training
Full framework training runs in notebooks or scripts on dedicated GPUs.
PyTorch · TensorFlow · JupyterLab
On the roadmap
Further out
Bigger jobs that need more than a single interactive session — we’re building toward them.
Serve LLM APIs
Expose a private inference endpoint for your app with vLLM or TGI.
vLLM · TGI
Large-model & multi-GPU training
Fine-tune bigger models as high-VRAM and multi-GPU hosts come online.
Multi-GPU hosts
Per-second billing
Pay only for the seconds you run.
Priced in rupees
UPI, cards, netbanking — no forex.
Sandboxed hosts
Isolated sessions on curated, hardened images.
India-based
Low-latency GPUs close to you.