AI Briefing#slm
The latest AI and developer news about #slm, with the original source and a short summary.
The latest AI and developer news about #slm, with the original source and a short summary.
A new method specializes Small Language Model vocabularies to reduce measured inference costs by 77% while improving quality scores on domain-specific tasks.
The dataset covers 106 SLMs with 150M parameters or fewer, evaluated on a Runpod B200 under Apache 2.0 license.
Mimir, a new 1.7B-scale model supporting English and Danish, has been released, demonstrating performance that surpasses Qwen and Gemma.
A Reasoning Corpus dataset with 5 million entries designed to improve the reasoning performance of small language models (SLMs) has been released.
LG AI Research has proposed the EXAONE LMM architecture, which resolves data licensing issues in existing LMMs while improving efficiency.
JetBrains has unveiled the Mellum2 model, a 12B-2.5A scale model optimized for fast inference.
Microsoft is testing the ability to run the Phi Silica model not only on Copilot+ PC's NPU but also on Nvidia RTX GPUs.
VibeThinker-3B, a small language model (SLM), has demonstrated frontier-level performance on math and coding benchmarks.
SupraLabs has released a 50M-scale reasoning model that generates a thinking process using a synthetic dataset.
Nex-AGI has unveiled Nex-N2-mini, a small language model for efficient inference.
Microsoft has announced two Aion 1.0 models optimized for on-device environments.
SupraLabs has released Supra-50M, a small language model with 50M parameters trained on 20 billion tokens.
A Claude Code skill automates everything from production traces to SLM deployment.
By specializing 3B and 7B open-source SLMs with SFT+DPO, they achieved higher scores than commercial models.
Introduces how to efficiently fine-tune LLMs using coding agents by combining Unsloth with Hugging Face Jobs.
In 2025, the deciding factor in enterprise AI was system design, not model performance.
IBM unveiled the ultra-compact Granite 4.0 Nano model series, optimized for on-device and edge environments.
SyGra, a low-code framework supporting data generation, transformation, and alignment for LLM and SLM training, has been released.
Upstage and LG Electronics have signed a memorandum of understanding (MoU) to implement on-device AI by embedding an SLM in LG gram laptops.
Kanana Nano-2.1B maintains 93% of Kanana Essence's Korean performance despite a 26% size reduction.
GCP's 5th generation Xeon (C4) instances demonstrated overwhelming efficiency in text embedding and generation performance compared to the previous generation.
Upstage and LG Electronics will jointly develop an on-device AI lightweight language model for LG gram.
Upstage's Solar 10.7B took first place on the Hugging Face Open LLM Leaderboard.