Search: “open-weights”

Today we're announcing OpenDocRouter: every model for document parsing under one API. There are ~4,000 OCR models on Hugging Face, and the f…

model-releasesjerry-liu--x

Today we're announcing OpenDocRouter: every model for document parsing under one API. There are ~4,000 OCR models on Hugging Face, and the frontier labs ship a new one nearly every month. Whatever's b

7 Oct 2026

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Research

Tree Navigation Without LLM Summaries: A Matched-Cost Study of Hierarchical Retrieval for Long-Document QA

DGX agent

arXiv:2610.06902v1 Announce Type: new Abstract: Retrieval-augmented generation grounds language models in external context, but for long documents flat top-k retrieval can cluster on a single region a

researcharxiv-cs-cl

7 Oct 2026

Weight Oracles: Reading Neural Network Weights with Language Models

arXiv:2610.07334v1 Announce Type: new Abstract: Interpretability methods for neural networks are predominantly reactive: they analyse activations produced during specific forward passes, requiring kno

7 Oct 2026

Artificial Analysis says Mistral Large 4 is the most intelligent model from outside the US and China, achieving results comparable to DeepSeek V4.1 Flash (max) (Artificial Analysis)

Artificial Analysis: Artificial Analysis says Mistral Large 4 is the most intelligent model from outside the US and China, achieving results comparable to DeepSeek V4.1 Flash (max) — Mistral has relea

model-releasestechmeme

ColdDDI: Evaluating Knowledge Utilization in Cold-Start Drug-Drug Interaction Prediction

arXiv:2610.05590v1 Announce Type: new Abstract: Cold-start drug-drug interaction (DDI) prediction tests whether models can identify clinically significant interactions for drugs without training-time

Industry

GLM-5.3's open release has yet to produce major attacks despite Anthropic's warnings about its Mythos-level cyber risk, undercutting calls to ban open models (Nathan Lambert/Interconnects AI)

Nathan Lambert / Interconnects AI: GLM-5.3's open release has yet to produce major attacks despite Anthropic's warnings about its Mythos-level cyber risk, undercutting calls to ban open models — Open-

industrytechmeme

Hidden in the Comments: A Context-Injection Attack Surface in Code LLMs

arXiv:2610.05139v1 Announce Type: new Abstract: Code large language model (Code LLM) assistants generate code from heterogeneous development contexts, including open files, imported modules, pasted sn

researcharxiv-cs-lg

How Does Geometry Enter Generated Motion?

arXiv:2610.05135v1 Announce Type: new Abstract: Under a fixed physical law, the visible geometry of a scene determines how motion must change. We ask how video generators realize this relationship. We

researcharxiv-cs-cv

InteractionBench: A Real-Time Interaction Benchmark for Streaming Video Systems

model-releasesarxiv-cs-cv

arXiv:2610.05775v1 Announce Type: new Abstract: A video assistant must speak when its instruction warrants a response and stay silent otherwise. We introduce a benchmark that evaluates this decision f

Introducing Mistral Large 4: Le chonk

model-releasessimon-willison

Introducing Mistral Large 4: Le chonk Mistral are back in the game. Today they're releasing a preview of Mistral Large 4, a 1 trillion parameter, 49 billion active parameter model trained on their own

Mistral launches a preview of Mistral Large 4, or Le Chonk, a 1T model it claims tops any open model developed in the US or Europe; weights are due October 27 (Carl Franzen/VentureBeat)

Carl Franzen / VentureBeat: Mistral launches a preview of Mistral Large 4, or Le Chonk, a 1T model it claims tops any open model developed in the US or Europe; weights are due October 27 — Mistral is

model-releasestechmeme

Mistral says it trained ML4 'from scratch' using 3,800 Nvidia Grace Blackwell GPUs in its data centers in Europe and much of its training data was multilingual (Mistral Blog)

Mistral Blog: Mistral says it trained ML4 “from scratch” using 3,800 Nvidia Grace Blackwell GPUs in its data centers in Europe and much of its training data was multilingual — Le Chonk — Today, we're

model-releasestechmeme

pplx-decider-v1.1-27b, our updated open weights multimodal decision model, is now available. It scores the highest on the new @huggingface D…

model-releasesperplexity--x

pplx-decider-v1.1-27b, our updated open weights multimodal decision model, is now available. It scores the highest on the new @huggingface Decision Index 0.3 benchmark. pplx-decider-v1.1-27b costs hal

Proxy Confidence: Auditing Black-Box LLM Agents with a Surrogate's Log-Probabilities

arXiv:2610.03894v1 Announce Type: cross Abstract: A deployed LLM agent emits tool calls, queries, and code that can be silently wrong -- by the time the error surfaces, the action has run. Frontier ch

agentsarxiv-cs-lg

Representational Control over Self-Report & Behavior Coherence in LLM Risk-Taking

arXiv:2610.04125v1 Announce Type: cross Abstract: Self-report is an appealing low-cost probe of an LLM's dispositions, but recent work finds only selective agreement between what models report and how

agentsarxiv-cs-lg

Same Pieces, Different Servers: A Tetris Benchmark for AI Agents as Served

arXiv:2603.02348v2 Announce Type: replace Abstract: An agent meets a model as served: through an endpoint with a price card, a shared cache and other tenants, or on whatever hardware a self-hosted mod

That one took some groundwork. Trained and served on our own compute, and RL shows no sign of saturation

model-releasesarthur-mensch--x

That one took some groundwork. Trained and served on our own compute, and RL shows no sign of saturation Meet Mistral Large 4, aka Le Chonk. • 1T parameters, natively multimodal. 49B active. It is the

We’re so back.

model-releaseslm-studio--x

We’re so back. Meet Mistral Large 4, aka Le Chonk. • 1T parameters, natively multimodal. 49B active. It is the best open weights model from US or Europe on aggregated benchmarks. • State-of-the-art on

A benchmark for LLMs playing Civilization V. GLM-5.3 is ahead of Opus-5.5, and Qwen-3.8-27B holds up surprisingly well.

model-releasesr-localllama

A while ago, I posted here getting OSS-120B and GLM-4.6 playing full games of Civilization V. Since then, models have moved pretty far, and we wanted a better understanding about models' capabilities

Agens Volundr 32B Preview: our small team's first model on our own hybrid architecture. Only 18 of 72 layers keep a KV cache (Apache-2.0)

model-releasesr-localllama

Hi r/LocalLLaMA. I'm on the team at Blockway, a small team in Hong Kong (disclosure: this is our model). Today we released Agens Volundr 32B Preview, the first model built on our own hybrid architectu

Benchmarking Literature Retrieval for a Model Organism: A Dictyostelium Case Study

model-releasesarxiv-cs-cl

arXiv:2610.03130v1 Announce Type: cross Abstract: Biological literature retrieval systems are often developed and evaluated using broad biomedical corpora and general-purpose search tasks. However, ma

Clinical Concept Centers in LLMs

arXiv:2610.02829v1 Announce Type: new Abstract: Large language models are increasingly used in clinical settings. However, research into the reliability and performance of these models has focused alm

researcharxiv-cs-cl

Counterexample Generation via Per-Theorem Symbolic Verifiers: When Imitation Hurts and Reinforcement Repairs

arXiv:2610.02444v1 Announce Type: cross Abstract: Large language models often solve a theorem forward yet fail to disprove a closely related false one: a falsification gap that supervised fine-tuning

Do Large Language Models Know Colombian Law? A Reliability Benchmark for the Colombian Legal System

arXiv:2610.03639v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used to support legal practice, education, and research, yet their reliability in national legal systems o

DyadMem: A Long-Term Memory Benchmark of How Agents Work with Users

arXiv:2610.03020v1 Announce Type: new Abstract: Long-term agents must remember not only what is true about a user, but also how a particular agent should work with that user as their shared history ev

Safety

Evaluating LLM-as-a-Judge Beyond Score Alignment: A Psychometric Analysis of Residual Judging Difficulty

arXiv:2610.02877v1 Announce Type: new Abstract: Large language models (LLMs) are widely used as automatic judges, with validity typically assessed via alignment with human scores. However, aggregate a

safetyarxiv-cs-cl

Fast Models, Slow Evidence: A Paired and Self-Audited Evaluation of System-1 Decision Models for LLM Agent Harnesses

arXiv:2610.02267v1 Announce Type: new Abstract: Agent harnesses make many small, typed decisions per task: which model to call, which tool to use, whether retrieved text is relevant, whether an input

agentsarxiv-cs-ai

Hardware

Ghost, which makes a 3,499 computer designed for AI agents and includes an RTX Pro 4000 SFF Blackwell GPU, emerges from stealth with an 11M seed led by a16z (Dominic-Madori Davis/TechCrunch)

Dominic-Madori Davis / TechCrunch: Ghost, which makes a 3,499 computer designed for AI agents and includes an RTX Pro 4000 SFF Blackwell GPU, emerges from stealth with an 11M seed led by a16z — For mu

hardwaretechmeme

GISTBench: Evaluating LLM User Understanding via Evidence-Based Interest Verification

arXiv:2603.29112v2 Announce Type: replace Abstract: We introduce GISTBench, a benchmark for evaluating Large Language Models' (LLMs) ability to understand users from their interaction histories in rec

How Robust Is Multimodal Claim Verification to LLM Rewriting?

arXiv:2610.02841v1 Announce Type: new Abstract: LLMs are known to introduce stylistic changes into generated text, yet how these stylistic shifts affect model decisions on scientific tasks remains und

researcharxiv-cs-cl

Improving the Energy-Efficiency of the Code Generated by LLMs through Effective Prompting

arXiv:2610.02571v1 Announce Type: cross Abstract: As AI-assisted programming becomes increasingly mainstream, the environmental impact of AI-generated software has emerged as an important consideratio

Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active. - Frontier reasoning efficiency - Advance…

Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active. - Frontier reasoning efficiency - Advances the Western open frontier on coding & agentic tasks - Trai

agentselon-musk--x

Safety

Labels Override Definitions in Jev-Style Typed Decision Models

arXiv:2610.02586v1 Announce Type: new Abstract: A typed decision model answers a fixed question about an input by returning a probability for each of several caller-defined options. Each option carrie

safetyarxiv-cs-ai

Let's go https://huggingface.co/reflection!

Let's go https://huggingface.co/reflection! Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active. - Frontier reasoning efficiency - Advances the Western op

agentsclem-delangue--x

MLCommons Jailbreak Benchmark v1.0

arXiv:2610.02827v1 Announce Type: new Abstract: Modern AI systems are designed to refuse hazardous requests. A jailbreak is a prompt crafted to bypass those safeguards and elicit outputs that the syst

Safety

OPD Before RL: Warm-Starting Rubric-Based RL with On-Policy Distillation

arXiv:2610.02781v1 Announce Type: cross Abstract: Many useful language-model tasks cannot be evaluated by exact outcome verification. Rubric-based reinforcement learning (RL) addresses this issue by s

safetyarxiv-cs-ai

Peer Influence across Heterogeneous AI Models

arXiv:2610.03095v1 Announce Type: new Abstract: When two AI agents disagree, who persuades whom? As multi-agent systems increasingly combine language models of different families and sizes, the answer

agentsarxiv-cs-ai

Prompted to Discriminate: Generalizing Malicious-Input Probes in the Wild

arXiv:2610.02413v1 Announce Type: new Abstract: LLM agents increasingly rely on activation probes as runtime monitors for prompt injection, jailbreaks, and unsafe requests, reading the model's own hid

Reflection unveils Beam, an open model it says rivals GLM 5.2 on reasoning while using 3x-4× less compute and approaches Qwen3.8-Max on coding and agentic tasks (Semafor)

Semafor: Reflection unveils Beam, an open model it says rivals GLM 5.2 on reasoning while using 3x-4× less compute and approaches Qwen3.8-Max on coding and agentic tasks — Reflection AI, the startup b

agentstechmeme

RxnOptBench: Benchmarking LLMs for Reaction-Condition Optimization in Organic Methodology

arXiv:2610.02242v1 Announce Type: cross Abstract: Chemical reaction-condition optimization -- choosing the catalyst, ligand, solvent, reagent, temperature, time, and atmosphere that jointly maximize y

Science Is Falling Behind the Frontier: Foundation Model Adoption Across Half a Million Papers

arXiv:2511.21739v2 Announce Type: replace-cross Abstract: We present the first large-scale analysis of AI foundation model usage in science -- not just citations or keywords. We find that adoption has

SyntaxBench: A Statistical Diagnostic Framework for Character-Level Reasoning in Large Language Models

arXiv:2610.03329v1 Announce Type: cross Abstract: Large language models are increasingly used where small syntactic errors matter, yet character-level reasoning is still evaluated mostly through isola

Thanks for sharing! We open-sourced code and models. 🤗Huggingface Paper: https://huggingface.co/papers/2609.38597 💻 Code: https://github.c…

model-releasesclem-delangue--x

Thanks for sharing! We open-sourced code and models. 🤗Huggingface Paper: https://huggingface.co/papers/2609.38597 💻 Code: https://github.com/nv-tlabs/PixelUMM 🤗 Model: https://huggingface.co/nvidia/Pi

Nonobench: an open benchmark of 49 LLMs on nonogram puzzles, public and open source [P]

model-releasesr-machinelearning

Nonobench measures how well LLMs solve nonograms (picross). Each model gets the row and column clues once and returns the full grid. No tools, one attempt per puzzle. Method: - Standard mode: 30 puzzl

4 Oct 2026

Hardware

Q&A with OpenAI VP of Hardware Richard Ho on its Jalapeño inference chip co-designed with Broadcom, using internal OpenAI models to design the chip, and more (Dr. Ian Cutress/More Than Moore)

Dr. Ian Cutress / More Than Moore: Q&A with OpenAI VP of Hardware Richard Ho on its Jalapeño inference chip co-designed with Broadcom, using internal OpenAI models to design the chip, and more — VP Ha

hardwaretechmeme

4 Oct 2026

A Comparative Explainability Framework for DeBERTa-v3 in Zero-Shot Medical Abstract Classification

arXiv:2610.02116v1 Announce Type: new Abstract: A comparative explainability framework is presented to audit DeBERTa-v3 under zero-shot classification of medical abstracts. The work addresses the disa

researcharxiv-cs-ai

2 Oct 2026

Architecture Without an Architect? Global Governance of Artificial Intelligence in a Divided World

arXiv:2610.01716v1 Announce Type: cross Abstract: Artificial intelligence presents an unusually difficult problem for global governance. The technology develops rapidly, crosses borders easily, and is

2 Oct 2026

Argo-Bench: Evaluating Data Agents on Enterprise-Scale Workflows