Nanthasit/sakthai-context-7b-merged

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7.6BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 6, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

Nanthasit/sakthai-context-7b-merged is a 7 billion parameter language model developed by Nanthasit, built upon Qwen/Qwen2.5-7B-Instruct. This merged checkpoint combines advanced tool-use capabilities with an extended 128k context length, making it suitable for complex agent-style tasks requiring structured outputs. It excels at open-ended text generation, tool-calling, and long-context processing, particularly for offline CPU/edge deployments.

Loading preview...

Model Overview

Nanthasit/sakthai-context-7b-merged is a 7 billion parameter language model developed by Nanthasit, continuing from Qwen/Qwen2.5-7B-Instruct. This model is a merged checkpoint, combining the tool-use behavior from Nanthasit/sakthai-context-7b-tools with the long-context capability from Nanthasit/sakthai-context-7b-128k.

Key Capabilities

  • Tool-Calling and Function-Calling: Designed for prompts requiring tool-use, utilizing <tools> XML prompt formatting for optimal performance.
  • Extended Context Length: Features long-context processing, making it suitable for agent-style task completion with structured outputs, with reliable performance up to ~8k–32k tokens.
  • Open-ended Text Generation: Capable of general text generation and chat for English use cases.
  • Offline Deployment: Intended for efficient deployment on offline CPU/edge devices via llama.cpp or Ollama.

Intended Use Cases

  • Agentic Workflows: Ideal for tasks requiring complex multi-step reasoning and interaction with external tools.
  • Long-Context Applications: Suitable for processing and generating content based on extensive input, such as document summarization or detailed conversational agents.
  • Research: Provides a foundation for research into small-to-mid tool-calling models under an MIT license.

Limitations

As a merged/continued checkpoint, users should verify its behavior before production use. Tool calling performance is sensitive to prompt formatting, and long-context generations are most reliable within specific token windows. Benchmarking against held-out tools is recommended prior to deployment.

Finetunes
1 models