Tag
VibeThinker-3B is a 3B-parameter model that achieves frontier-level reasoning performance on math, coding, and STEM benchmarks by optimizing the Spectrum-to-Signal Principle (SSP) post-training pipeline, reaching performance comparable to much larger models.
This paper introduces IAPO, a reinforcement learning algorithm that improves tool-calling capabilities in multimodal small language models by aligning input attribution with a stronger teacher. Experiments on Qwen2.5-VL-3B show an average 3% improvement in visual question answering accuracy across six test sets.
The author shares learnings from training a 160M parameter LLM from scratch, experimenting with architectures like multi-token prediction and hierarchical reasoning models. They emphasize the importance of fast iteration, simplifying ideas, and understanding why architectures work.
Microsoft announced two new on-device AI models at Build 2026: Aion 1.0 Instruct, an open-weights small language model, and Aion 1.0 Plan, a 14B parameter reasoning and tool-calling model for local agentic workflows.
Trained a 75M parameter LLM called KeyLM from scratch on 18B tokens, achieving competitive instruction-following scores against larger models while using fewer parameters and less data.
OpenBMB releases MiniCPM5-1B, a leading 1B open weights LLM that achieves the highest Artificial Analysis Intelligence Index score (17.9) in its size class, surpassing larger models like Qwen3.5 2B while using fewer parameters.
MiniCPM5-1B is a new 1B parameter AI model from OpenBMB featuring hybrid reasoning with Think/No-Think modes, 128K context, and Apache 2.0 license, running on various hardware.
MiniCPM5-1B is fully open-sourced with weights, training data, and deployment code; it achieves top scores among sub-2B models and runs on edge devices.
SupraLabs released Supra-50M, a compact 50M-parameter causal language model with base and instruct versions, trained on 20B tokens from fineweb-edu, achieving competitive benchmarks against larger models like GPT-2 and SmolLM.
Sapient Intelligence introduces HRM-Text, a 1B-parameter reasoning language model trained on only 40B tokens with a budget of $1,000, achieving competitive performance while drastically reducing data and compute requirements.
Microsoft released Fara-7B, a 7-billion parameter small language model that can autonomously control a computer to perform tasks like clicking, scrolling, and filling forms, running on-device and beating larger models like OpenAI's computer-use agent on benchmarks.
Microsoft released Fara-7B, a 7 billion parameter agentic small language model for computer use, achieving state-of-the-art performance among models of its size and competitive with larger systems.
Cactus-Compute released Needle, a 26M parameter open-source model distilled from Gemini for efficient on-device function calling using a novel Simple Attention Network architecture without MLPs.
Two new open-source small language models are being released: one matches state-of-the-art accuracy at up to 93x smaller size, and the other outperforms a recent OpenAI model. The first model drops tomorrow.
OpenBMB has released MiniCPM V4.6, a 1B-parameter multimodal large language model optimized for mobile devices under the Apache 2.0 license. It features mixed visual token compression and claims approximately 1.5x faster throughput than Qwen3.5 0.8B while running natively on iOS, Android, and HarmonyOS.
DR-Venus-4B is a 4B-parameter deep-research agent trained on only 10K open samples via agentic SFT+RL with turn-level rewards, outrunning prior sub-9B agents and rivaling 30B models on research benchmarks while staying deployable on edge devices.
Z-lab releases DFlash for Qwen3.6-35B-A3B, a model fine-tuning/compression technique, with training complete and weights now available on GitHub and HuggingFace.
Cactus-Compute releases Needle, a 26M parameter distilled model from Gemini 3.1, using a pure attention architecture optimized for on-device inference and local fine-tuning.
Microsoft released Fara-7B, an efficient 7 billion parameter agentic small language model (SLM) for computer use tasks, achieving state-of-the-art performance within its size class and competitive with larger systems.
Hugging Face released SmolVLM2-256M-Video-Instruct, a lightweight open-source multimodal model that handles video, image, and text inputs while requiring only 1.38GB of GPU RAM, making it well-suited for on-device applications and domain fine-tuning.