---

library_name: gguf

license: apache-2.0

pipeline_tag: text-generation

quantized_by: Abiray

base_model: InternScience/Agents-A1

tags:

• MoE
• agent
• tool-use
• long-horizon

---

Agents-A1-Q4_K_M-GGUF

This repository contains the GGUF quantized version of Agents-A1, a 35B Mixture-of-Experts (MoE) agentic model developed by InternScience. Quantization has been performed using the Q4_K_M method to optimize performance and reduce memory consumption while preserving agentic reasoning capabilities.

• Original Model: InternScience/Agents-A1
• Technical Report: arXiv:2606.30616

---

Provided Files

| File Name | Size | Quantization Method | Description |

| :--- | :--- | :--- | :--- |

| Agents-A1-Q4_K_M.gguf | 21.2 GB | Q4_K_M | Medium 4-bit quantization. Recommended balance between size, speed, and accuracy. |

---

About Agents-A1

Agents‑A1 is a 35B Mixture‑of‑Experts agentic model built to scale heterogeneous agentic abilities across multiple domains including Long‑horizon Search, Engineering, Scientific Research, Instruction Following, and Tool-calling.

Key Highlights

• Agentic Reasoning: Excels at decomposing complex tasks into executable sub-steps, planning ahead, and adapting its strategy based on intermediate results.
• Tool Use: Natively supports function calling and tool integration, enabling seamless interaction with APIs, code interpreters, search engines, and external tools.
• Scientific and Professional Reasoning: Handles tool-integrated scientific reasoning and professional knowledge question answering.
• Instruction Following: Precisely follows detailed, multi-constraint instructions across diverse domains.

---

Performance Overview

Despite operating in the ~35B model class, Agents-A1 delivers highly competitive performance against frontier-scale systems such as GPT-5.5, DeepSeek-V4-pro, and Kimi-K2.6. It achieves overall SOTA results on several challenging benchmarks, including Seal-0 (56.4), HiPhO (46.4), FrontierScience-Olympiad (79.0), FrontierScience-Research (40.0), IFBench (80.6), and IFEval (94.8).

---

Deployment & Usage

You can run this GGUF model using popular inference engines like llama.cpp or Ollama.

Option 1: Using llama.cpp

To run inference via the command line with llama.cpp, download the file and use the following command structure:

./llama-cli \
  --model Agents-A1-Q4_K_M.gguf \
  --prompt "<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n<|im_start|>user\nHello, who are you?<|im_end|>\n<|im_start|>assistant\n" \
  --ctx-size 8192 \
  --threads 8