miromind-ai/MiroThinker-v1.5-30B

Overview

MiroThinker-v1.5-30B is an open-source research agent model designed to advance long-horizon reasoning and complex multi-step analysis. It supports extended context windows and deep agent-environment interactions, positioning it for demanding tasks such as extensive reasoning workflows and information-seeking applications. The model is released in the 30B parameter scale with notable capabilities for tool integration and reasoning depth.

Official model page:
https://huggingface.co/miromind-ai/MiroThinker-v1.5-30B

Support for this model on DeployPad is planned and will be made available soon.


Launch

### Launch
[Launch on DeployPad](https://deploypad.geodd.io/launch/miromind-ai/mirothinker-v1.5-30b){: .btn .btn-primary}


Supported Hardware (Planned)

Support is planned for the following GPUs:

GPU
NVIDIA H100
NVIDIA H200
NVIDIA RTX Pro 6000 (Blackwell)

Model Characteristics

Attribute Details
Model Family MiroThinker v1.5
Parameter Scale 30B
Context Window Up to 256K tokens
Tool Calls Support Up to 400 per task
Primary Focus Long-horizon reasoning and research workflows
Architecture Interactive agent-oriented model

Key capabilities include long-context reasoning, deep multi-step analysis, and enhanced interaction depth with environments, making it suited for tasks where extended sequences and reasoning chains are important.


Supported Configurations (Planned)

Capability Status
Precision FP8 (planned)
Streaming Inference Planned
OpenAI-compatible API Planned
Dynamic Batching Planned
GPU Sharing / Multi-tenant Planned
Production Deployment Planned

Performance Notes

Support for MiroThinker-v1.5-30B will enable deployment across the listed GPUs with optimized inference configurations. Official benchmarks are not yet available for DeployPad, and performance characteristics will be published once testing is complete.


Current Status

State
Coming Soon

Feedback Requested

To help refine performance profiles and deployment guidance, users are encouraged to share feedback on:

  • Output behaviour and reasoning quality

  • Deployment stability across hardware targets

  • Any errors or unexpected behaviours

Feedback can be posted in this thread or under Support & Troubleshooting.