# Nanbeige4.2-3B: Unlocking Agentic Capabilities in a Compact Model

arXiv: https://arxiv.org/abs/2607.22083 · HTML: https://mlsanity.com/paper/2607.22083/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- Placed in General LLM agent frameworks and benchmarks, one of 16 directions in LLM agents and tool use.
- Builds on 28 of its 34 references that are in the map.
- First posted on arXiv 24 Jul 2026.

## Placed in

- LLM agents and tool use / [General LLM agent frameworks and benchmarks](https://mlsanity.com/agents/general-llm-agent-frameworks-and-benchmarks/)

## Builds on (most cited first, 28 of 28)

- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Program Synthesis with Large Language Models](https://mlsanity.com/paper/2108.07732/) (2021)
- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark](https://mlsanity.com/paper/2406.01574/) (2024)
- [Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them](https://mlsanity.com/paper/2210.09261/) (2022)
- [Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces](https://mlsanity.com/paper/2601.11868/) (2026)
- [SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?](https://mlsanity.com/paper/2509.16941/) (2025)
- [Gemma 4 Technical Report](https://mlsanity.com/paper/2607.02770/) (2026)
- [Generalizing Verifiable Instruction Following](https://mlsanity.com/paper/2507.02833/) (2025)
- [SciCode: A Research Coding Benchmark Curated by Scientists](https://mlsanity.com/paper/2407.13168/) (2024)
- [Scaling Latent Reasoning via Looped Language Models](https://mlsanity.com/paper/2510.25741/) (2025)
- [GDPval: Evaluating AI Model Performance on Real-World Economically Valuable Tasks](https://mlsanity.com/paper/2510.04374/) (2025)
- [Beyond Benchmarks: MathArena as an Evaluation Platform for Mathematics with LLMs](https://mlsanity.com/paper/2605.00674/) (2026)
- [Mixture-of-Recursions: Learning Dynamic Recursive Depths for Adaptive Token-Level Computation](https://mlsanity.com/paper/2507.10524/) (2025)
- [ResearchRubrics: A Benchmark of Prompts and Rubrics For Evaluating Deep Research Agents](https://mlsanity.com/paper/2511.07685/) (2025)
- [Towards Robust Mathematical Reasoning](https://mlsanity.com/paper/2511.01846/) (2025)
- [MCP-Atlas: A Large-Scale Benchmark for Tool-Use Competency with Real MCP Servers](https://mlsanity.com/paper/2602.00933/) (2026)
- [OfficeQA Pro: An Enterprise Benchmark for End-to-End Grounded Reasoning](https://mlsanity.com/paper/2603.08655/) (2026)
- [DeepResearch Bench II: Diagnosing Deep Research Agents via Rubrics from Expert Reports](https://mlsanity.com/paper/2601.08536/) (2026)
- [Tmax: A simple recipe for terminal agents](https://mlsanity.com/paper/2606.23321/) (2026)
- [ToolMind Technical Report: A Large-Scale, Reasoning-Enhanced Tool-Use Dataset](https://mlsanity.com/paper/2511.15718/) (2025)
- [ClawGym: A Scalable Framework for Building Effective Claw Agents](https://mlsanity.com/paper/2604.26904/) (2026)
- [AgentIF-OneDay: A Task-level Instruction-Following Benchmark for General AI Agents in Daily Scenarios](https://mlsanity.com/paper/2601.20613/) (2026)
- [Nanbeige4-3B Technical Report: Exploring the Frontier of Small Language Models](https://mlsanity.com/paper/2512.06266/) (2025)
- [Nanbeige4.1-3B: A Small General Model that Reasons, Aligns, and Acts](https://mlsanity.com/paper/2602.13367/) (2026)
- [IQuest-Coder-V1 Technical Report](https://mlsanity.com/paper/2603.16733/) (2026)
- [VibeThinker-3B: Exploring the Frontier of Verifiable Reasoning in Small Language Models](https://mlsanity.com/paper/2606.16140/) (2026)

## Built on by (most cited first, 11 of 11)

- [NeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing Harness](https://mlsanity.com/paper/2609.08183/) (2026)
- [LoopSpec: Pipelined Self-Speculative Decoding for Looped Transformers](https://mlsanity.com/paper/2609.17184/) (2026)
- [Quantizing Looped Transformers: Feedback Exposure and Calibration Blindness](https://mlsanity.com/paper/2609.30820/) (2026)
- [Nanbeige4.2-3B on Apple Silicon: Fixing Deployment Bugs and Decreasing Looped Transformer Memory Overhead](https://mlsanity.com/paper/2608.13987/) (2026)
- [WaveFront Decoding: Parallelized Self-Speculative Decoding for Looped Language Models](https://mlsanity.com/paper/2609.23033/) (2026)
- [Looping Beyond Twice: A Scalable Recipe for Looped Mixture-of-Experts](https://mlsanity.com/paper/2610.01153/) (2026)
- [Universal interpolation for deep residual self-attention networks](https://mlsanity.com/paper/2610.01981/) (2026)
- [ALoDLM: Adaptively Looped Diffusion Language Models](https://mlsanity.com/paper/2610.04198/) (2026)
- [Enabling Dynamic Computation in Looped LMs](https://mlsanity.com/paper/2610.09013/) (2026)
- [From Chain-of-Thought to Loops: Non-Autoregressive Latent Reasoning via Looped Transformers](https://mlsanity.com/paper/2610.11472/) (2026)
- [LoopQuant: Accurate and Efficient Post-Training Quantization for Looped Transformers](https://mlsanity.com/paper/2610.16368/) (2026)

## Abstract (arXiv)

We present Nanbeige4.2-3B, a compact general agentic model with 3B non-embedding parameters. It delivers strong performance across code-agent, office-agent, and complex tool-use tasks while maintaining highly competitive reasoning capabilities in mathematics, coding, and science. Nanbeige4.2-3B is pretrained from scratch on 28T tokens with a Looped Transformer that reuses the layer stack to increase capacity without adding parameters. For SFT data and trajectory construction, we expand the diversity of executable environments, task assets, and agentic scaffolds through real-world deployment and large-scale synthesis. Our RL pipeline applies mixed-mode RLHF over Think and Non-Think responses to improve overall model quality and reduce failure cases, length-controlled reasoning RL to balance accuracy and reasoning efficiency, and agentic RL with outcome and process rewards to stabilize long-horizon training. Extensive evaluations show that Nanbeige4.2-3B outperforms larger models, including Qwen3.5-9B and Gemma4-12B, across diverse agentic benchmarks while remaining competitive on reasoning and alignment tasks. Performance with OpenClaw further supports its use as a compact local personal assistant.