Python SDK for LoRA fine-tuning with supervised fine-tuning and reinforcement learning
Linux sandboxes for Claude Code, Codex and Cursor over MCP protocol
Stage
Not given
Not given
Founded
Not given
Not given
Based in
Not given
Not given
Pricing model
Not given
Paid
Pricing
Not given
Plans from $10 to $1,000 per month. Credit lands each month and boxes bill by the hour: Nano $10/month ($0.014/hour), Micro $25/month ($0.034/hour), Extra Small $50/month ($0.068/hour), Small $100/month ($0.137/hour), Medium $250/month ($0.342/hour), Large $500/month ($0.685/hour), Extra Large $1,00
Key features
forward_backward() operation for computing loss and accumulating gradients
optim_step() for updating the adapter
sample() for generating from current or saved policy
save_state() to save and resume training state
Tinker API compatibility
Separate trainer and sampler resources
Synchronous and asynchronous RL patterns
Frozen base model weights with LoRA adapter updates
Support for open models from Qwen, Google, DeepSeek, Moonshot AI, NVIDIA, and Z.ai
Sign up without API key or credit card
Fork running sandboxes to run work in parallel
Pause and resume sandboxes
Pre-configured images with Python, Node.js and git
Upload and download files between project and sandbox
Idle sandboxes auto-pause after 15 minutes
Paused sandboxes auto-delete after 7 days
What makes it different
Managed training and sampling infrastructure, no setup required
Tinker-compatible API for easy migration
Ready-made SFT and RL recipes through Loops cookbook
Support for vision-language models
No API key required for initial sign-up
Integrates directly with Claude Code, Codex and Cursor via MCP protocol
Unused monthly credit rolls over up to two months
Pay only for compute time; sleeping boxes bill only for storage