reallexi/lexi-rm-agent
TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 22, 2026License:otherArchitecture:Transformer0.0K Featherless Exclusive Cold
The reallexi/lexi-rm-agent is a 495 million parameter language model developed by Reallexi LLC, derived from Qwen/Qwen2.5-0.5B-Instruct. It is specifically fine-tuned for customer support chatbot interactions, leveraging the Bitext customer support training dataset. This model is optimized for efficient deployment with its merged adapter, requiring no PEFT at runtime, and operates with a trained context length of 512 tokens.
Loading preview...
reallexi/lexi-rm-agent Overview
The reallexi/lexi-rm-agent is a compact yet capable language model developed by Reallexi LLC, featuring 495 million parameters. It is built upon the Qwen/Qwen2.5-0.5B-Instruct base model and has been extensively fine-tuned using the bitext/Bitext-customer-support-llm-chatbot-training-dataset.
Key Capabilities & Features
- Optimized for Customer Support: Specifically trained on a large customer support dataset, making it highly suitable for chatbot applications in this domain.
- Efficient Deployment: The model's adapter has been merged into the base weights, eliminating the need for PEFT (Parameter-Efficient Fine-Tuning) at runtime, which simplifies deployment and reduces overhead.
- Compact Size: With 495M parameters, it offers a balance between performance and resource efficiency, making it suitable for environments with memory constraints.
- Memory Footprint: Requires approximately 944 MB for FP16/BF16 weights, 472 MB for 8-bit, and 260 MB for 4-bit quantization.
- Training Details: Underwent 5 epochs of training, learning from 100,000 samples over 1,250 steps using an Auto LoRA strategy with a rank of 8 and alpha of 16.
Good For
- Customer Service Chatbots: Ideal for developing responsive and context-aware chatbots to handle customer inquiries and support tasks.
- Resource-Constrained Environments: Its relatively small size and merged adapter make it a strong candidate for deployment where computational resources or memory are limited.
- Quick Integration: The simplified deployment due to the merged adapter allows for faster integration into existing systems.