Search: “open-weights”

arXiv:2605.25676v1 Announce Type: new Abstract: We release Llamion, a family of 14B-parameter open-weight language models obtained by transforming Orion-14B into the standardized Llama-family architec

arXiv:2605.15613v1 Announce Type: new Abstract: Through an analysis of sequences generated by open-weight large language models (LLMs), we demonstrate that LLMs are heavily biased toward English. Whil

Are scaling laws finally working for time series foundation models? Today, @datadoghq is releasing Toto 2.0 weights in Apache 2.0 on @huggingface. It's a family of open-weights TSFMs from 4M to 2.5B p

A little talk on what we can learn from implementing LLM architectures from scratch in Python and PyTorch. And how I approach new open-weight models, compare them against reference implementations etc

arXiv:2510.05497v5 Announce Type: replace-cross Abstract: Large-scale Mixture of Experts (MoE) Large Language Models (LLMs) have recently become the frontier open-weight models, achieving remarkable m

arXiv:2605.10777v1 Announce Type: new Abstract: The quality of open-weight language models has dramatically improved in recent years. Sharing weights greatly facilitates model adoption by enabling the

arXiv:2605.04177v1 Announce Type: new Abstract: As LLMs enter conflict monitoring, understanding systematic distortions in their outputs is critical for humanitarian accountability. We evaluate four v

arXiv:2604.01496v2 Announce Type: replace-cross Abstract: We introduce SWE-ZERO to SWE-HERO, a two-stage SFT recipe that achieves state-of-the-art results on SWE-bench by distilling open-weight fronti

arXiv:2605.03052v1 Announce Type: new Abstract: We study how Large Language Models (LLMs) process negation mechanistically. First, we establish that even though open-weight models often provide wrong

Today we're releasing ZAYA1-8B, a reasoning MoE trained on @AMD and optimized for intelligence density. With <1B active params, it outperforms open-weight models many times its size on math and reason

Want to run the same harness across multiple interfaces? Try ACP. Deep Agents ships with it out of the box. open-weight LLMs have come a long way on agent tasks! but the harness you wrap them in matte

arXiv:2603.08182v2 Announce Type: replace-cross Abstract: Large language models often underperform in many European languages due to the dominance of English and a few high-resource languages in train

arXiv:2603.25268v2 Announce Type: replace Abstract: We introduce CRAFT, a multi-agent benchmark for evaluating pragmatic communication in large language models under strict partial information. In thi

arXiv:2604.25119v1 Announce Type: new Abstract: Auditing the fine-tunes of open-weight generative models for harmful specialization has become a new governance challenge for model hosting platforms. T

arXiv:2604.23801v1 Announce Type: new Abstract: Practitioners deploying small open-weight large language models (LLMs) for medical question answering face a recurring design choice: invest in a domain

arXiv:2604.24346v1 Announce Type: cross Abstract: Vision-language models (VLMs) are increasingly deployed as evaluators in tasks requiring nuanced image understanding, yet their reliability in scoring

FLUX 2 Dev is an open-weight, 32-billion-parameter AI model developed by Black Forest Labs for text-to-image generation and advanced image editing . The model is available in ComfyUI and Diffusers fra

Kimi K2.6 has achieved the top ranking on OpenRouter's weekly LLM leaderboard, marking a significant achievement for the model. The announcement expresses gratitude to developers using Kimi and indica

DeepSeek released a new open-weights model with strong benchmark performance, though Mollick notes that benchmark results may not fully capture the capabilities of open models compared to closed syste

This article demonstrates running Gemma 4, Google's open-weight language model, on NVIDIA's Jetson Orin Nano Super edge computing device. It likely covers the model's capabilities, performance metrics

Kimi K2.6, a model developed by Moonshot AI, has achieved the top ranking on OpenRouter's programming leaderboard, indicating superior performance in code generation and programming-related tasks comp

arXiv:2508.00161v3 Announce Type: replace-cross Abstract: The releases of powerful open-weight large language models (LLMs) are often not accompanied by access to their full training data. Existing in

arXiv:2604.18510v1 Announce Type: cross Abstract: Open-weight language models can be rendered unsafe through several distinct interventions, but the resulting models may differ substantially in capabi

📢 Kimi K2.6 API is live • Input Price (Cache Hit): 0.16 / M tokens • Input Price (Cache Miss): 0.95 / M tokens • Output: $4.00 / M tokens Kimi K2.6 is our latest + most intelligent model - stronger lo

Kimi K2.6 autonomously overhauled exchange-core, an 8-year-old open-source financial matching engine. Over a 13-hour execution, the model iterated through 12 optimization strategies, initiating over 1

arXiv:2604.15075v1 Announce Type: cross Abstract: Open-weight Small Language Models(SLMs) can provide faster local inference at lower financial cost, but may not achieve the same performance level as

Qwen 3.6 Plus Preview is Alibaba's next-generation large language model released on March 30-31, 2026 , and the first open-weight variant was released following the February Qwen 3.5 series, prioritiz

A mid-2026 outlook piece from the Interconnects AI newsletter in which the author makes specific predictions about the trajectory of open-weight language models, likely covering expected capability mi

As of April 2026, Latent Space published a curated ranking of the top locally-runnable AI models, providing a research-based overview of the best open-weight or self-hostable models available at that

arXiv:2604.08644v1 Announce Type: new Abstract: This technical report introduces EXAONE 4.5, the first open-weight vision language model released by LG AI Research. EXAONE 4.5 is architected by integr

Ollama announced a meetup focused on Gemma, Google's open-weight language model family, scheduled for Wednesday. The event likely brought together developers and AI enthusiasts to explore running Gemm

MiniMax-M1 is a large-scale open-weight reasoning model developed by MiniMax, featuring a hybrid Mixture-of-Experts (MoE) architecture with 456 billion total parameters (activating 45.9 billion per to

Google released Gemma 4 in April 2025 as its most capable open-weight model family to date, built on the same research as Gemini 3 and licensed under Apache 2.0 for unrestricted commercial use, fin...

Yesterday’s Mythos announcement from Anthropic was overblown. • Sandboxing was turned off, so test didn’t show much about the real world. • Cheap open-weight models can (already) do some similar stuff

'But here is what we found when we tested: We took the specific vulnerabilities Anthropic showcases in their announcement, isolated the relevant code, and ran them through small, cheap, open-weights m

Google released Gemma 4 on April 2, 2026 — a family of four open-weight models built on the same research as Gemini 3 and licensed under the permissive Apache 2.0 license, marking a significant shi...

On October 16, 2026, China’s Ministry of Commerce and Ministry of Science and Technology released a draft revision to the Catalog of Technologies Prohibited and Restricted from Export. It proposes add

arXiv:2610.14438v2 Announce Type: replace Abstract: Large Language Models (LLMs) are increasingly being used to summarize and verify claims in critical sources that inform consequential decisions. Suc

arXiv:2610.19830v1 Announce Type: cross Abstract: Cyber Threat Intelligence (CTI) investigators who monitor dark web marketplaces for leaked credentials and breach listings must pass the diverse CAPTC

arXiv:2610.14346v2 Announce Type: replace Abstract: Enterprise customer-support agents increasingly rely on Large Language Models (LLMs) to execute bounded, tool-using workflows. We study this problem

arXiv:2604.06138v2 Announce Type: replace-cross Abstract: Long-context audio reasoning is underserved in both training data and evaluation. Existing benchmarks target short-context tasks, and the open

Meta’s Muse Spark Open, a high‑capacity US open‑weight model for agentic software engineering, is now live on Together AI as a Day 0 launch partner. The model supports a 1‑million‑token context window

TaoLive released a new TaoMate H3 FL2VA checkpoint, so I converted it for ComfyUI and added it to the same Hugging Face repo as my previous TaoMate conversion. The new file is BF16, around 1.24 GB ins

Meta’s Muse Spark Open is an 816‑billion‑parameter mixture‑of‑experts coding model that activates only 42 billion parameters per token through top‑8 routing across 256 experts, and has been trained on

What makes NVIDIA Nemotron 3.5 Super VL different? → Text, image, document, and video understanding in one model → Reasoning and tool use with visual context preserved across steps → 120B hybrid MoE (

arXiv:2610.15666v1 Announce Type: new Abstract: Converting natural-language instructions into executable action plans conditioned on a scene image is a core capability for embodied human-robot interac