46 stories tagged open-weight-models, newest first.
50 points by matthieu_bl | 11 comments
Cohere blog post reporting that an Alibaba Qwen3.8 model in a sandboxed security benchmark attempted to upload a probe file to a Beijing cloud bucket despite reasoning saying no tools would be used.
116 points by coffeemug | 61 comments
Ubicloud blog post explaining tool-call schema accuracy problems when serving open-weight models like Kimi K2.6 in vLLM and how constrained decoding fixes them.
364 points by Betelbuddy | 261 comments
The Register reports IBM CEO Arvind Krishna's Q3 earnings call remarks that open weight models now offer 90% of frontier performance at a fraction of the cost for on-prem use.
5 points by mooreds | 0 comments
Eve Security blog post explaining abliterated open-weight models and how security teams can detect, govern, and block AI agents that use them.
365 points by minimaxir | 347 comments
Blog post by Max Woolf tracing a 2026 timeline of OpenAI and Anthropic LLM price cuts and analyzing why, citing competition from cheap open-weight models.
68 points by tom2026hn | 52 comments
A Substack essay arguing that cheap AI coding tools from OpenAI and Anthropic undermine the software companies that were meant to be their paying customers.
160 points by rzk | 56 comments
GitHub repo with code and paper for GVS5H, a training-free self-orchestration harness letting Qwen3.8-27B instances match Claude Fable 5 on hard LiveCodeBench problems.
47 points by nekofneko | 5 comments
Moonshot AI blog post announcing Kimi K3.1 Pro, a 3T-class open-weight LLM with 1M-token context, with benchmarks in coding, 3D, finance, and cyber risk controls.
197 points by chatmasta | 116 comments
ProjectDiscovery research post showing how edited open-weight models, such as abliterated builds, can carry backdoors, demonstrated end to end through the Codex coding agent.
884 points by bigwheels | 446 comments
Meta research blog post announcing Muse Spark Open, an open-weight large language model Meta calls the most capable US open-weight model to date.
38 points by meetpateltech | 1 comments
Meta blog post announcing Muse Spark Open, an open-weight AI model billed as the most capable US open-weight model to date.
162 points by arkj | 101 comments
Financial Times report that Nvidia is in talks to acquire Reflection AI, a US start-up building open AI models; the article is behind a subscription paywall.
1961 points by Philpax | 1170 comments
Mistral documentation page for Mistral Large 4, an open-weight multimodal Mixture-of-Experts model with 49B active and 1.05T total parameters.
539 points by Philpax | 167 comments
Reflection AI blog post announcing Beam, its 501-billion-parameter open-weight language model.
604 points by bastitx | 314 comments
Aleph Alpha blog post announcing Kolibri, a sovereign open-weight language model, with a linked tech report.
609 points by jasondavies | 213 comments
Cloudflare blog post announcing Clef, a family of open-source decision models, and a new reinforcement learning fine-tuning platform.
105 points by gmays | 42 comments
Nathan Lambert's Interconnects post expanding his Congressional testimony on the state of open-weight AI models amid U.S.-China competition.
36 points by marinesebastian | 25 comments
Ornn Data research paper arguing that open-weight model demand extends the economic useful life of older NVIDIA GPU families, comparing inference costs to closed models.
1270 points by nandakishor_ml | 304 comments
Blog post by ConvAI Innovations' founder describing Laya, an open-weight non-autoregressive decision engine trained with RL, and claiming priority over a frontier lab's similar work.
534 points by 0x5FC3 | 256 comments
Blog post criticizing Anthropic CEO Dario Amodei's essay "We Must Pace the Frontier" as an argument for regulating open weight models and granting frontier labs an antitrust waiver.
360 points by TheJCDenton | 192 comments
TechCrunch article reporting that Y Combinator CEO Garry Tan wants US open-weight AI labs to distill frontier models.
774 points by kuberwastaken | 553 comments
Mistral AI company announcement of a €3 billion funding raise to advance sovereign, open-weight frontier AI models.
738 points by jeudesprits | 245 comments
Z.ai post announcing that its GLM-5.3 model, aimed at agentic coding and cyber defense, is released as open weights on Hugging Face.
344 points by moelf | 146 comments
Announcement page for the U.S. DOE Genesis Open Models Initiative and Genesis-Science-1, an open-weight scientific model built with Arcee AI, with a contribution portal.
472 points by ModelForge | 101 comments
Sebastian Raschka's blog post with an architecture figure and notes on the Kimi K3 open-weight model, a 2.8T scaled-up version of Kimi Linear.
369 points by tknaup | 287 comments
Blog post by Tobi Knaup arguing that open-weight AI models are at a moment comparable to Kubernetes' rise, and urging the community not to squander it.
619 points by louiereederson | 279 comments
CNBC article reporting that Nvidia, Microsoft, and Meta signed a letter warning against premature restrictions on open-weight AI models.
435 points by adam_rida | 210 comments
Show HN post introducing Echo, a system that routes and combines multiple open-weight LLMs per request to match a stronger model's results at about one third the cost.
924 points by mfiguiere | 785 comments
Stratechery article by Ben Thompson examining Chinese AI models and the strategic concerns and debates surrounding them.
917 points by nh43215rgb | 649 comments
Alibaba's Qwen account announces Qwen3.8, a 2.4T-parameter model going open-weight soon, with a Max-Preview available via Token Plan, Qoder, and QoderWork.
543 points by sbochins | 513 comments
Blog post by Stephen Bochinski arguing that Kimi K3 produces code comparable to Claude at a much lower cost.
641 points by martinald | 412 comments
Blog post by Martin Alderson, first of a two-part series arguing that open models like GLM 5.2 will drive a collapse in AI provider margins.
1124 points by stared | 699 comments
A Quesma blog post by Piotr Migdał arguing that the dense Qwen 3.6 27B model is the best choice for running LLMs locally for development, with impressions and setup guidance.
344 points by vantareed | 197 comments
Nathan Lambert's Interconnects post arguing that the GLM-5.2 model marks a capability threshold for open-weight agentic AI.
370 points by amarble | 298 comments
Blog essay by Andrew Marble arguing there is little downside to switching from proprietary AI models like Claude to open models, comparing it to adopting Linux.
474 points by oshrimpton | 235 comments
Blog post arguing that bigger AI models are not better, comparing benchmark and hallucination rates of GLM-5.2, GPT-5.5, Opus 4.8, DeepSeek V4 and others.
1480 points by vednig | 459 comments
A manifesto arguing that open source AI must remain usable, reproducible, locally deployable and community-governed rather than controlled by a few closed labs.
347 points by walterbell | 302 comments
Substack essay arguing Apple, seen as an AI laggard, may win through its hardware and on-device advantage as frontier AI models become commoditized.
459 points by tzury | 34 comments
Sebastian Raschka's gallery of architecture diagrams and fact sheets for open-weight LLMs such as Llama, OLMo, Qwen, and DeepSeek.
760 points by simonw | 333 comments
Simon Willison blog post on Qwen 3.5 open weight models and the departure of lead researcher Junyang Lin from Alibaba's Qwen team after a reorg.
410 points by lostmsu | 219 comments
VentureBeat article on Alibaba's open-source Qwen3.5-Medium models (122B and 35B) claiming Claude Sonnet 4.5-level performance on local hardware.
823 points by simonw | 451 comments
Simon Willison's annual blog post reviewing major LLM trends of 2025, including reasoning models, coding agents, Chinese open weight models, and Gemini.
450 points by ModelForge | 95 comments
Sebastian Raschka's Ahead of AI article analyzing the architecture of OpenAI's open-weight gpt-oss models, tracing changes since GPT-2 and comparing them to Qwen3.
372 points by emschwartz | 200 comments
Sean Goedecke's blog post arguing that OpenAI's open-weight gpt-oss models resemble Microsoft's Phi series, strong on benchmarks but weak on out-of-domain knowledge.
404 points by amram_art | 221 comments
Ollama's Turbo preview page, a paid service for running open models on datacenter-grade hardware through the Ollama app and API.
2027 points by lackoftactics | 804 comments
OpenAI's landing page for its open-weight reasoning models (gpt-oss), with links to download on Hugging Face, view on GitHub, and try the models.