Model Lineup

Weaver supports a variety of open-weight models for fine-tuning.

Supported Models

Nex-AGI Optimized Models

These are optimized versions of base models, fine-tuned for enhanced reasoning and instruction-following capabilities.

Model ID	Parameters	Type	Context Length
nex-agi/Qwen3-30B-A3B-Nex-N1	30B (3B active)	MoE	128K
nex-agi/Qwen3-32B-Nex-N1	32B	Dense	128K
nex-agi/DeepSeek-V3.1-Nex-N1	671B (37B active)	MoE	128K

Qwen Series

Model ID	Parameters	Type	Context Length
Qwen/Qwen3-8B	8B	Dense	128K
Qwen/Qwen3-32B	32B	Dense	128K
Qwen/Qwen3-30B-A3B	30B (3B active)	MoE	128K
Qwen/Qwen3-235B-A22B	235B (22B active)	MoE	128K

DeepSeek Series

Model ID	Parameters	Type	Context Length
deepseek-ai/DeepSeek-V3.1	671B (37B active)	MoE	128K
deepseek-ai/DeepSeek-V3.2	671B (37B active)	MoE	128K

Model Types

• Dense: Standard transformer architecture with all parameters active
• MoE (Mixture of Experts): Only a subset of parameters are active per token, enabling larger models with similar computational cost

Choosing a Model

For agent scenarios:

• Prioritize Nex-AGI optimized models (nex-agi/* series) for superior reasoning and tool-use capabilities
• These models are specifically fine-tuned for agentic workflows and multi-step problem solving

For other applications:

• Start with Qwen/Qwen3-8B or Qwen/Qwen3-32B for balanced performance
• deepseek-ai/DeepSeek-V3.1 for best quality
• Use Nex-AGI optimized versions for enhanced reasoning capabilities
• MoE models offer strong performance with efficient inference

Usage

To use any model, specify its Model ID when creating a training client:

from weaver import ServiceClient

service_client = ServiceClient()
training_client = service_client.create_model(
    base_model="Qwen/Qwen3-8B",  # Replace with your chosen model
    lora_config={"rank": 32}
)

Next Steps

• Learn about Training and Sampling - Core APIs
• Explore Loss Functions - Available losses
• Understand Saving and Loading - Model persistence