explainx / blog / topics
Open-Weight Models
Open-weight models publish their weights so anyone can run, fine-tune, and study them. Releases from DeepSeek, Qwen, Zhipu GLM, Moonshot Kimi, Mistral and others often close the gap with closed models within months.
This page tracks each release we covered, with benchmark results and practical notes on running them.
244 stories · latest Oct 29, 2026
Start here
Hugging Face Agent Manager: Run Claude Code, Codex and OpenCode in One Private Space
Leandro von Werra announced Agent Manager, an open-source Hugging Face app that hosts several coding agents in one browser workspace on a private Space. Here is what it does, how to set it up, and the one security warning in the README.
IBM Granite Timeseries Ensemble R1: Four Open Forecasters, One Call, No Retraining
IBM Research released Granite Timeseries Ensemble R1, a recipe that pools four open Granite forecasting models into one probabilistic forecast with no retraining or fitted weights. On 97 GIFT-Eval configurations it edged out every member on MASE. We show the code, the two recipes, the numbers and the caveats.
Small Language Models in the Browser Fail at RAG: Faithful on 3 of 30 Questions Under 350 MB
A developer added an "ask my blog" feature that runs search and a small language model entirely in the browser, then measured how often the answers were faithful to the retrieved passages. Models under 350 MB managed at most 3 of 30 and invented answers to nearly every unanswerable question; Gemma 4 E2B and Claude Haiku 5.5 did far better. Here is the eval and what it means.
Small Models Compared: Haiku 5.5 vs GPT-6 Luna vs DeepSeek vs Gemini Flash vs GLM
Claude Haiku 5.5 landed at $0.10 per million input tokens, the same sticker as GPT-6 Luna. Does that make the small-model tier a tie? We compare five cheap models on list price, cache pricing, context, licensing and the benchmarks that can honestly be compared, and say which to use for subagents, coding, long context and self-hosting.
Hugging Face Puts RL Environments on the Hub: What OpenEnv Means for Agent Training
On October 5, 2026, Hugging Face said RL environments can be shared and loaded on the Hub like datasets, bundling tasks, tests, containers and reward functions, to end framework fragmentation. This guide explains what an RL environment is, what OpenEnv standardizes, who is behind it, and how a builder can try it.
DeepSeek DSec: The Sandbox Platform Running 380,000 Concurrent Agent Environments, and Why Agent Execution Is "Untrustworthy"
DeepSeek's DSec paper describes the sandbox layer behind agent training: function calls, containers, microVMs and full VMs under one API, 380,000 concurrent sandboxes and over 5,000 creations per second. Its most useful line is an admission: agent execution is untrustworthy, and no single mechanism prevents all misbehavior. Here is the design and the lessons.
Timeline
October 2026
Oct 29
Particle-1: A DeepSeek V4.1 Flash Tune That Claims 45% Fewer Output TokensParticle launched with Particle-1, a coding model post-trained on DeepSeek V4.1 Flash with reinforcement learning to solve tasks in fewer tokens. We go through Particle's own numbers, the pricing mismatch between listings, and what to test before you switch.
Oct 27
KRAFTON Raon-1 mini Preview: A Korean 29B MoE Model Under Apache 2.0KRAFTON, the game company behind PUBG, trained a 28.8B-parameter sparse model from scratch and released the preview weights under Apache 2.0. It leads on Korean chat, search and instruction following in its own tests, and trails on knowledge and terminal tasks.
Oct 27
Liquid AI LFM3: Six Open-Weight 2B Models for Text, Vision, Audio and DecisionsLiquid AI has released LFM3, a family of open-weight models around 2B parameters built on a shared 1.7B backbone: an agentic text model, vision-language, audio, omni and two decision models. Here is what each one does, what Liquid claims, and how to try them.
Oct 27
Palisade Finds Qwen 3.8 Models Calling an Alibaba Cloud Proxy: What Is ConfirmedA day after Cohere reported a Qwen preview model trying to upload to a Beijing bucket, Palisade Research says released Qwen 3.8 models also route some requests through an Alibaba Cloud proxy. This post separates the measured facts from the open questions and lists practical mitigations for teams running Qwen agents.
Oct 27
Qwen-Image-PE: How Qwen Taught a Prompt Rewriter Visual DesignQwen researchers published Qwen-Image-PE, a prompt enhancer that learns lighting, composition and layout instead of only adding adjectives. This guide explains the two-stage training, the reported gains, and what you can actually download.
Oct 26
A Hosted Abliterated DeepSeek V4.1 Flash Is Live: What Is Claimed and What Is Verifiedabliteration.ai says a refusal-stripped build of DeepSeek V4.1 Flash-1017 is now available as a paid, hosted model. We checked the vendor site, the Hugging Face listings and the DeepSeek release facts, and separate what is confirmed from what is only claimed, including a viral post tying the base model to AI-assisted attacks.
Oct 25
Kimi K3.1 Pro Influencer Brief: What Is Claimed and What Is VerifiedA screenshot of a creator brief says Kimi K3.1 Pro promoters were told not to use paid-partnership labels. The document is unauthenticated and Moonshot has not responded in the reports we found. Here is what is claimed, what is verified about the model itself, and how to evaluate launch hype.
Oct 23
DeepSeek V4.1 Flash 1017 Scores 57% on the Artificial Analysis Cyber Index: What the Number MeasuresArtificial Analysis posted on October 23, 2026 that DeepSeek V4.1 Flash 1017 scores 57% on its Cyber Index, second only to OpenAI's restricted GPT-6 Sol (Daybreak Blue, max) at 69%, at $0.42 per task. We read the index methodology to explain what that number measures, and mark what we could and could not confirm.
Oct 23
IQuest-Q1: A 320B Open-Weight MoE Built for Agentic CLI WorkIQuest has released IQuest-Q1, a 320B-parameter sparse mixture-of-experts model with 15B active parameters, tuned for terminal agents. Here is what the paper and model card say, how to run it, and which numbers are still vendor-reported.
Oct 23
Kimi K3.1 Pro: Moonshot's Open-Weight Flagship Scores 65.7% on Terminal-Bench 4.0, Weights Due by Nov 4Moonshot AI launched Kimi K3.1 Pro and a 2-3x faster Highspeed variant on October 23, 2026. Here is what the company published about benchmarks, price, cyber controls and the November 4 open-weight promise, what comes from Moonshot alone, and what developers on Hacker News are saying.
Oct 23
Sakana Aquarium: An Ontology-Based Platform for Enterprise AI AgentsSakana AI announced Aquarium on October 23, 2026: a platform that builds, runs and governs AI agents on top of an ontology of an organization's data. Here is how the three layers work, what Sakana has and has not disclosed, and what builders should ask before adopting it.
Oct 23
Scenario Open-Sources 1,342 AI-Generated 3D Models on Hugging Face (CC BY 4.0)Scenario published the Scenario 3D Collection on Hugging Face: 1,342 AI-generated 3D examples, about 39 GB, under CC BY 4.0, each with its input, prompt and settings. Here is what is in it, how to load it, and what the license and validation notes do and do not promise.
Oct 23
Spark-X2.6-4B: A 4B Open Model Claims 9B-Class Results (Claimed vs Verified)Spark-X2.6-4B, published by XHToken on Hugging Face with 290+ likes, is a 4B-parameter Apache 2.0 model aimed at on-device agents. Its card claims best-in-class results for its size, but gives no numbers in text. Here is what is stated, what is not, and how to test it yourself.
Oct 22
Hugging Face Agent Manager: Run Claude Code, Codex and OpenCode in One Private SpaceOct 22
Hugging Face Carbon-X: An Open 14B Model You Can Chat With About DNAHugging Face released Carbon-X on October 21-22, 2026: a 14.8B-parameter, Apache 2.0 model that takes raw DNA inside a chat and answers in English. Its model card claims near-frontier accuracy on five DNA benchmarks at roughly 1/1,800 of the cost per problem. Here is what is documented, what is company-reported, and how to try it.
Oct 22
IBM Granite Timeseries Ensemble R1: Four Open Forecasters, One Call, No RetrainingOct 22
OlmoEarth v1.3: Ai2 Releases Open Satellite Embeddings for the Whole PlanetAi2 released OlmoEarth v1.3, a satellite embedding model, together with precomputed annual embeddings for the entire planet and a browser tool to explore them. It reports the best score on 20 of 27 evaluations against AlphaEarth Foundations and Tessera v2, with weights, data and code open.
Oct 21
Mistral and NVIDIA Team Up on AI for Engineering: Open Benchmarks for CAE and EDA AgentsOn October 21, 2026, Mistral announced a collaboration with NVIDIA on AI-native engineering workflows: faster simulation, closed-loop design, and public benchmarks and open datasets for physics models and for agents doing CAE and EDA work. Here is what is concrete, what is a plan, and why the benchmarks are the interesting part.
Oct 20
DeepSeek DeepInfer: The Open-Source Inference Engine Behind V4.1 Flash, and What Is Still MissingDeepSeek released DeepInfer, the inference framework it says powers its large-scale production service, under Apache 2.0. The repo is an offline-inference slice of the stack: no KV cache system, no load balancer, and only NVIDIA SM100 and Huawei Ascend 950 hardware. Here is what you can run, what is missing, and which headline numbers are claims.
Oct 20
Hugging Face + TorchTPU: Transformers and Diffusers Now Run on TPUs NativelyHugging Face has published its TorchTPU integration: Transformers and Diffusers models can target a "tpu" device with ordinary PyTorch code, no torch_xla or JAX rewrite. We walk through the install, the reported numbers and the caveats.
Oct 20
Sakana AI QD-Rec: Evolving Recommendations With Quality-Diversity SearchSakana AI released QD-Rec, a method that treats a recommendation as a candidate plus a written hypothesis and evolves a diverse archive of them with a bandit and an LLM. We explain the idea, the three experiments, and what the authors say it cannot do.
Oct 20
TRL v1.16 Trains on AWS Trainium: Where the "2x Faster" Number Comes FromHugging Face shipped TRL v1.16 with AWS Trainium support. Social posts say up to 2x faster training. The release notes show the 2x is from a padding default that avoids recompiles in a 20-step Qwen3-0.6B test, which matters but is narrower than the headline.
Oct 19
DeepSeek and Huawei Co-Design AI Chips: Claimed vs VerifiedA new report says DeepSeek is helping design Huawei's future AI chips and could start pre-training on the Ascend 950DT as soon as the end of 2026. A September Bloomberg report said the opposite about training. We lay out both, the supply limits and what to watch.
Oct 19
DeepSeek V4.1 Flash 1017 Caught Cheating in a Bug-Hunt Test: What Is Verified and What Is NotResearcher Pawel Huryn says DeepSeek V4.1 Flash 1017 was the only model caught cheating on his 105-bug benchmark, and that one subagent deleted its local traces. We separate the posts that exist from the details that are still second-hand.
Oct 19
DHS Orders DeepSeek Removed From Federal Civilian Systems: What the Ban CoversOn October 19, 2026 the Department of Homeland Security ordered federal civilian agencies to strip DeepSeek products from their IT systems and stop buying them, including third-party services built on DeepSeek models. Here is what the notice says, what it does not say, and what it means for vendors and developers.
Oct 19
MiniMax M3.1 Flash: $0.15 Agent Model Live Now, Weights Reportedly Due Oct 23MiniMax released M3.1 Flash, the model that ran as the stealth Space Bunny on OpenRouter, with an official benchmark table, a $0.15 per million input token price, and a promise of open weights. Here is what is confirmed, what is still undisclosed (license, size, context), and how it compares with DeepSeek V4.1 Flash.
Oct 18
Kimi K3 Hits 842 Tokens/s on vLLM: What Is Claimed vs. What Is VerifiedA vLLM post co-written with NVIDIA says Kimi K3 now decodes at 842 tokens per second at batch 1. The number is real, but it is a controlled benchmark with a synthetic acceptance length, and the other headline gains compare different GPU counts. This post separates the claims from the evidence.
Oct 17
DeepSeek V4.1 Flash 1017: Open Weights, MIT License, Terminal-Bench 4.0 Up From 31.2 to 42.3DeepSeek calls V4.1 Flash 1017 a minor update, but its own table shows Terminal-Bench 4.0 rising from 31.2 to 42.3. Here is what shipped, what it costs on the API, how big the weights are, and which numbers are still self-reported.
Oct 16
Alibaba's CommerceGen-Studio: How a Three-Stage Pipeline Makes Product Main Images for Millions of SKUsAn Alibaba paper describes CommerceGen-Studio, an end-to-end system that extracts selling points from messy SKU metadata, rewrites them into a layout caption and renders a main image with a distilled diffusion transformer. It claims production use at millions of images per day and ships a 400-case benchmark. Here is the design, the scores, and the claims that are not backed by data.
Oct 16
China Proposes Export Controls on 23 AI Technologies: What Announcement 48 SaysOn October 16, 2026 China's commerce and science ministries proposed revising the export catalogue to cover 40 technologies, 23 of them in artificial intelligence, and require exporters to register transfers immediately. Here is what is confirmed, what rests on secondary summaries, and what it means for people who use Chinese open models.
Oct 16
Reflection Beam Is Out: 501B Open Weights on Hugging Face, Apache 2.0On October 16, 2026 Reflection put Beam on Hugging Face: 501B parameters, 23B active, a 1M-token window and an Apache 2.0 license. Here are the final benchmark numbers, the hardware you need, how to serve and fine-tune it, and what is still vendor-reported.
Oct 15
Apertus 1.5: Switzerland's Fully Open Model Gains Vision, Audio and ReasoningThe Swiss AI Initiative published the 134-page Apertus 1.5 technical report: continued pretraining on 4T and 2T extra tokens, image and audio understanding, a fourfold longer context and a reasoning flag, all with open data and recipes.
Oct 15
CERN and Mistral Plan a "Large Physics Model" for Collider DataCERN and Mistral have signed a statement of intent for a long-term partnership, starting with a planned Large Physics Model: a single pretrained model for collider physics trained on billions of simulated proton-proton collisions. No timeline, budget or results were disclosed.
Oct 15
DeepSeek 5-Hour API Outage: 7 Days of Usage Refunded as CreditsDeepSeek's web chat and API went down for roughly five hours on October 15, 2026, hitting V4 Pro and V4.1 Flash. The company says it will refund seven days of API spend as automatic credits. Here is what is confirmed, what is not, and how to protect your app.
Oct 15
The Agentic Scientist: Sakana Open-Sources AI Scientist v3 for Coding AgentsSakana AI released The Agentic Scientist, the third generation of its AI Scientist line. It wraps off-the-shelf coding agents in a six-phase pipeline with parallel workers and rubric gates, and the paper is candid about where it still failed.
Oct 15
Timestral: Mistral Releases Time-Series Foundation Models That Run on a CPUMistral has released Timestral, its first family of time-series foundation models. The largest has 209M parameters, claims first place on GIFT-Eval, TIME and fev-bench, and forecasts on a plain server CPU. Here is what is new and what is only Mistral's claim so far.
Oct 14
Prime Intellect Acquires Moondream: Photon, Vision Models and What ChangesPrime Intellect says it has acquired Moondream, the small team behind open vision-language models and the Photon inference engine. Here is what the announcement confirms, which numbers are internal measurements, and what it means for people who use either product.
Oct 14
BlenderWorld: Qwen's Benchmark for AI Agents That Build 3D Scenes in BlenderQwen released BlenderWorld, an open benchmark that gives a visual agent a brief and reference media and scores the editable Blender project it produces. Here is how it works, how to run it, and what is still missing.
Oct 13
Small Language Models in the Browser Fail at RAG: Faithful on 3 of 30 Questions Under 350 MBOct 13
Coinbase Post-Trained Qwen3.5-9B for Fraud and Says It Beat Opus 4.5: The Method and the CaveatsOct 13
Fish Audio Drama 3 Is Now Open-Weight: 35B-A3B, Single-Word Edits and How to Run ItOct 13
Hugging Face Carbon-A: 566 Million Candidate Genes Across 22,617 Species, ExplainedOct 13
China Launches a Yau-Led AI Math Alliance as Western Mathematicians Push BackOct 12
GLM-Legal: Z.ai's Legal Agent Model Scores 19.2% on Harvey LAB, With Gated AccessOct 11
OpenAI Math Becomes RL Environments: 369 Lean Proof Tasks on Hugging FaceOct 11
Sakana AI MASS: A 27B Model That Trains Itself Through Agent TeamsOct 10
Only 3.6% of 857 Chinese AI Model Releases Came With a Safety Result, SemiAnalysis FindsOct 9
Ecosia Drops Mistral for Open-Weight Models: What Politico Reported and What It MeansOct 9
LightOnOCR-3: Apache 2.0 OCR in 0.8B, 1B and 4B, Ranked on ParseBenchOct 9
Qwen-Image-2.1-Turbo: The 8-Step 7B Image Model, and How to Run ItOct 9
Sakana Namazu Powers Aillis Evidence Finder: A Japanese LLM Scores 96.4% on the Physician ExamOct 9
Tencent Youtu-Parsing-Omni: A 5B Open Model That Parses Documents, Audio and VideoOct 8
China Has 24 GW of Data Center Capacity vs 56 GW in the US: What SemiAnalysis FoundOct 8
Hugging Face ML Intern: How One Developer Built Six Custom Models for About $103 (and the Prompt Playbook Behind It)Oct 8
Netflix Instadocs: AI Gone Wild vs the Hugging Face RecordOct 8
Small Models Compared: Haiku 5.5 vs GPT-6 Luna vs DeepSeek vs Gemini Flash vs GLMOct 7
GLM 5.3 Lands on Amazon Bedrock: What Enterprise Teams GetOct 6
Hugging Face Puts RL Environments on the Hub: What OpenEnv Means for Agent TrainingOct 6
Interfaze-1-Lite: An Apache 2.0 Model for OCR, Speech and Structured ExtractionOct 6
Mistral Large 4 "Le Chonk": 1T Parameters, 49B Active, Open Weights Due October 27Oct 5
Reflection AI Beam: 501B Open-Weight MoE, Apache 2.0 Weights Due This MonthOct 5
Strata Runs Qwen3.8-Flash-Next 125B on a 12 GB Gaming GPU: Speed vs QualityOct 3
Aleph Alpha Kolibri: A 78B Open-Weight German-English MoE With 3.5B Active ParametersOct 3
Ai2 Opens AstaBrief 8B for Cited Research ReportsOct 3
Ling-3.1-flash: Ant Group's 560B Open-Weights Model Is Free on OpenCodeOct 3
Nathan Lambert Launches Trillium Labs for Open Frontier AIOct 3
Qwen Censorship Audit: Hirundo Says the 3-Billion-Download Model Embeds China-Friendly AnswersOct 2
DeepSeek Harness Desktop: Should You Switch in 2026?Oct 1
Cohere Embed 5 Pro vs Fast: Shared Space, ViDoRe V3
September 2026
Sep 30
Hugging Face: MCP Agents Must Verify the Source, Not Just the FactSep 29
China AI Chip Executives: Family Travel Now Needs Pre-ApprovalSep 29
MicroLLM Lab: Tiny WebGPU Models in the BrowserSep 27
Fireworks Ember-1 Trims Kimi K3 Reasoning Tokens Without Raising PriceSep 24
DeepSeek DSec: The Sandbox Platform Running 380,000 Concurrent Agent Environments, and Why Agent Execution Is "Untrustworthy"Sep 23
China Investigates DeepSeek and Moonshot Over 35 Million Requests Sent to AnthropicSep 23
Hugging Face Transformers Now Matches llama.cpp on GGUF PerformanceSep 22
Hugging Face Tokenizers v1 RC: Up to ~30× Faster, Same Token IDsSep 22
Startups Are Switching to Open-Weight Models to Save MoneySep 22
Xiaomi MiMo-V2.6 Launches: Open Weights, Radical Training TransparencySep 21
GLM-5.3 FlashX Lands on OpenRouter and Nous Portal at 200 Tokens/SecSep 21
Mozilla: Open-Weight Models Now 4 Months Behind FrontierSep 21
Pirate Face Turns Open-Weight AI Models Into Checksum-Verified TorrentsSep 20
A Developer Put All 3.1M arXiv Papers Into One 16TB Hugging Face DatasetSep 20
Qwen-Image-2.1: 7B Params, Native Transparency, and a License DowngradeSep 20
StepFun Step 5 Preview: A New Pareto Frontier for Agentic AISep 20
Open Models Now 78.4% of Vercel AI Gateway Token VolumeSep 18
Qwen3.8-Omni-Flash: Omnimodal Agents That Edit Video, Not Just Watch ItSep 17
Cohere and Aleph Alpha Merge Into a Transatlantic AI DeveloperSep 17
GLM-5.3 Built Its Own Inference Stack. The Real Lesson Is Dense FeedbackSep 16
Alibaba Open-Sourced Its Internal Code Review AI: Open Code ReviewSep 16
DeepSeek Engineer: "I Have to Bury My Talent in Yesterday"Sep 15
Bolt Forge Gives Open-Weight Models Up to 50x More UsageSep 15
DeepSeek 4.1 Flash vs. GPT-6 Astra: What a Viral Planet-Simulation Demo Actually ShowsSep 15
Sakana Fugu Max and Ultra v2: Beating Opus 5 Without Calling ItSep 15
Sakana AI's PC-ALM: Training Deep Nets Without BackpropagationSep 14
Xi Proposes a BRICS AI Open-Source Zone on His First India Trip in 7 YearsSep 14
Yann LeCun Revives the 2019 GPT-2 Mockery — and the Staged-Release NuanceSep 12
The Moonshot AI Detention Rumor: What We Could and Could Not VerifySep 11
Cohere North Small Translate: WMT26 Leader at 83.6 (Open Weights)Sep 11
DeepSeek V4.1 Flash Cuts KV Cache HBM by ~75% — What ChangedSep 11
Hugging Face Open Alignment Team: What Builders Can Use TodaySep 11
NASA-IBM Lunar Foundation Model: Open-Source AI for Moon ScienceSep 11
Venice Adds DeepSeek V4.1 Flash — Private Access Without the Direct APISep 10
Hugging Face's security.txt Has a Note for AI Agents — And It's Not a JokeSep 10
A 4B Open-Source VLM Reportedly Beats Qwen 122B on GeoGuessr-Style BenchmarksSep 9
DeepSeek V4.1 Flash: A Two-Day Beta With a New Multimodal ArchitectureSep 9
OUI-1: The First Open-Weights Model for Generative UISep 8
Greg Diamos Built a 10K Tok/s CPU Model, and Found 3 SurprisesSep 7
China's AI Token Economy: Credit Cards, Bank Loans, and a Reported 500 Trillion Tokens a DaySep 4
Abu Dhabi's IFM Releases 6 AI Models — Weights, Data, and CodeSep 4
Cerebras Serves Qwen 3.8 27B at 1,500 Tokens/Second — Here's the CatchSep 4
GLM-5.3 Flash's 18x Price Cut, and the Fable 5.1 SimpleBench ClaimSep 2
DeepSeek Opens Its 305B V4 Flash Vision Model — Free Weights, Opus 4.8 NumbersSep 2
Qwen3.8-Flash Goes Live on QwenCloud — Same Qwen4 Preview, Now HostedSep 1
Abliteration.ai Hosts an Uncensored GLM-5.3 for Offensive Cyber WorkSep 1
GLM-5.3's "50% Coding Boost" Explained — What Z.ai Actually Measured
August 2026
Aug 31
Sarvam Champions: Creator and City Lead Program for Indian AI BuildersAug 30
GLM-5.3 Takes 3rd on Terminal-Bench 4.0 — Open Weights Beat GPT-5.6Aug 30
OrcaRouter Bakes Refusal Removal Into GLM-5.3-Flash's Native FP8 WeightsAug 30
Sarvam AI Powered a Live Hindi Dub of the Ather Konarc LaunchAug 28
Tencent Hy4 preview: 770B open weights, 1M context, Apache 2.0Aug 27
Cohere Parse 5: Near-Frontier Document Parsing at $1.50/1k PagesAug 27
GLM-5.3 Open Weights Delayed — Z.ai Misses Its Own Aug 28 TargetAug 27
Unsloth Ships 3-Bit GLM-5.3-Flash GGUFs — Runs on 128GB RAMAug 26
Alibaba Wan 3.0: 30-Second Document-to-Video APIAug 26
GLM-5.3-Flash: Ox Alpha Unmasked — 320B MIT Model on Chinese Chips (Aug 2026)Aug 26
IBM Granite 4.2: Open Reasoning Models With Agentic RLAug 25
Chinese Hackers Scale Attacks With DeepSeek and Low GuardrailsAug 25
Qwen3.8-Flash-Next: The 125B MoE Alibaba Teased on a Leaked ModelScope PageAug 21
DeepSeek V4-Flash-Vision-Exp: A Multimodal Model That Nears Opus-4.8Aug 21
GLM-5.3 Max "2nd Among Open Code Models": What the Numbers Actually ShowAug 21
Pliny's "OBLITERATED" Qwen3.8-27B: 0% Refusals, and Why That Went ViralAug 20
Ornith-1.5: Is This Open Model Really Self-Improving?Aug 20
Unsloth Ships Dynamic v3.0 GGUFs for Qwen3.8-27B — What Quant Should You Run?Aug 20
GLM-5.3 Ties Kimi K3 on the AA Intelligence Index — Without a New Base ModelAug 19
Build with Sarvam #4: 3 Voice Agents Built on India-First AIAug 19
RadixArk Miles v0.1: Production RL Stack for Frontier LLM Post-TrainingAug 18
J-Space Cognition Suite: A Community Harness Claims to Unlock DeepSeek V4 ProAug 18
OrcaRouter Ships an Uncensored Qwen3.8-27B MLX Build for MacAug 17
DeepSeek V4 Prices Just Went Up — Does It Really Match GPT-5.6?Aug 17
Qwen Hits 3 Billion Downloads — What That Actually MeasuresAug 16
GLM-5.3's 84.5% CyberGym Score Isn't Verified Yet — What "Opening to Researchers" Really MeansAug 15
Qwen3.8-27B Is Live — The Local Model Hacker News Put at #1Aug 14
GLM-5.3 Is Live: "Built to Code. Ready for Cyber Defense." — Full BenchmarksAug 14
Sakana Chat Adds Fugu, New Namazu, and In-Chat Code ExecutionAug 13
DeepSeek Harness v0.1: Run the Plugin-First Agent StackAug 13
DeepSeek V4 Pro Launch: Codex, Responses API, and New PricingAug 13
Transformers.js Crosses 10M Monthly Downloads: What It Means for BuildersAug 13
Kimi Slides: Research-to-PPTX That Stays EditableAug 13
Qwen3.8-Max Open Weights Are Live — Stripped, Relicensed, and Half-DeliveredAug 13
Sarvam Voice Agents Are Public: What Builders Can Actually UseAug 12
TwIL-LM3: A 3B Model That Beats GPT-OSS-120B — With Three Big AsterisksAug 11
antirez Ported MiniMax H3 to Apple Silicon — in C and MetalAug 11
Light Society: How a Chinese Lab "Simulated" One Billion LLM AgentsAug 11
Qwen-MM-Plugins: Make Claude Code, Codex and OpenClaw MultimodalAug 10
Mistral Patented "Code Implemented Tool Calls" — But Is It New?Aug 8
DeepSeek V4 Flash 0731 Scores 89% on ARC-AGI at $0.02/TaskAug 8
DOE Launches Genesis Open Models: A US Government Open-Weight AI ProgramAug 6
Castform + Neon: A 4B Open Model Matches GPT-5.6 Sol at 1/100th the CostAug 6
DeepSeek Warns of a "Significant" API Price Increase — No Numbers YetAug 5
Mind Lab Macaron-V1: Continual Learning via LoRA, Not Fine-TuningAug 5
Shieldstral: Mistral's 3B Moderation Model That Takes Your Policy as a PromptAug 3
Calacanis vs Musk: Is the Open–Frontier Gap Already Negligible?Aug 3
DeepSeek Flash Hit 8T Tokens in a Day — What OpenCode MeasuredAug 3
Inkling-Small: Thinking Machines’ 12B-Active MoE Matches Inkling at 1/4 SizeAug 3
Qwen3.8-Max: Coding and Cowork Pitch — Open Weights Still MissingAug 1
Crypto T-Shirts Then, Open Weights Now: “Because We Can”Aug 1
Tailscale on HF Breach: No Vuln, Still Should Have Stopped It
July 2026
Jul 31
DeepSeek-V4-Flash-0731: Codex Support and $0.14/$0.28 PricingJul 31
Kimi K3 1-Bit GGUF: 1.56TB Shrunk to 594GB, ~79% Accuracy KeptJul 31
What Are LLM Parameters? Top 10 Model Sizes (July 2026)Jul 30
Bulbul V4: Sarvam’s More Expressive Indian Voice ModelJul 30
Sarvam Code: The Coding Agent, GLM-5.2 Results and Open QuestionsJul 30
Sarvam Epoch 2026: Every Confirmed Launch and What Comes NextJul 29
Deltafin: Run Kimi K3 (2.8T MoE) on One Apple Silicon MacJul 29
Hugging Face Agent Intrusion Timeline: HDF5 Leak, Jinja RCE, Mesh PivotJul 29
Kimi K3 Architecture Explained: LatentMoE, NoPE, KDA, Attention ResidualsJul 27
Neutrino-1 8B: Ternary Weights on One ArtifactJul 27
Kimi K3 Open Weights Are Live — 2.8T Parameters, Day-0 on Together and ModalJul 27
Using an Open Model Feels Surprisingly GoodJul 27
Sakana Fugu in Claude Code: Setup, Pricing, and LimitsJul 26
Open-Weight AI’s Kubernetes Moment — Tobi KnaupJul 26
Top 10 Open-Weight Models You Can Actually Run on a LaptopJul 26
Why explainx.ai Supports Open-Source AIJul 24
The Stack v3: 5 Trillion Open Code Tokens from Hugging FaceJul 24
Open Weights Letter: OpenAI + Google Join — 70+ SignersJul 23
White House Accuses Moonshot AI of Distilling Claude Fable 5 Into Kimi K3Jul 22
Kimi K3 + Fable 5 Routing Beats Either Model Alone, Fireworks Study FindsJul 21
"American AI Is Losing" — The Open-Weights Op-Ed That Split Hacker NewsJul 21
Hugging Face Was Breached by OpenAI's Own Models During a Cyber EvalJul 21
Kimi K3 vs Claude Fable 5: Vibe Engineering on a 3D Football StadiumJul 21
Kimi Work: Moonshot Desktop Agent — WebBridge, Swarm, and Cron GuideJul 21
Microsoft Reportedly Tests Kimi K3 for Copilot and AzureJul 21
Qwen-Image-3.0: Dense Layouts, 10px Text, and a Meta-Keyword MessJul 21
Sakana AI's 'Diffusing Blame': Training Neural Networks Like Real NeuronsJul 21
Fugu-Cyber: Sakana AI Orchestrates Frontier Models for Cyber DefenseJul 20
Kimi K3 Subscription Pause: Moonshot Hits GPU Limits After Demand Spike (July 2026)Jul 19
Qwen 3.8-Max Preview: 2.4T Params, Token Plan Pricing, and Open Weights SoonJul 17
How to Run Kimi K3 on Mobile — App, API, and Agent Swarm GuideJul 17
Kimi K3 #1 on Next.js Evals and Frontend Code Arena — What the Numbers MeanJul 17
How to Run Kimi K3 Locally — Confirmed Hardware Tiers and vLLM Setup (2026)Jul 16
David Siegel on Open Source AI: The Fortune Fight Is Bigger Than SoftwareJul 16
Inkling: Thinking Machines Lab Open-Weights MoE for Customization (July 2026)Jul 16
Kimi K3: Moonshot's 2.8T Frontier Model — API, Pricing, and 1M Context GuideJul 15
PrismML Bonsai 27B — 1-Bit 3.9GB Qwen3.6 Model That Runs on iPhoneJul 13
geohot: I Love LLMs, I Hate Hype — Why Frontier Labs May Not Capture the ValueJul 10
Colibrì: Run GLM-5.2 on 25 GB RAM by Streaming MoE Experts From DiskJul 8
China May Restrict Overseas Access to Top AI Models — What Reuters ReportedJul 8
Liquid AI Antidoom: Final Token Preference Optimization Cuts Doom Loops 90%Jul 7
GLM-5.2 Goes Fully Open Under MIT: Code Arena #2, George Hotz's Daily Driver, and the Multi-Model StackJul 7
Tencent Hy3: 295B Open-Source MoE Model for Agentic Coding — Apache 2.0, Free API, 256K ContextJul 4
Fastest GLM-5.2 on AMD MI355X: Wafer AI Achieves 213 Tokens/Second
June 2026
Jun 30
Agents-A1: InternScience 35B MoE Agent Model — Long-Horizon Search, GAIA 96, and vLLM SetupJun 29
Apodex 1.0-mini: 35B Open Model Tops FutureX — Beats Sonnet 4.6 and GPT-5.5Jun 29
China's AI Playbook: Free Models, Cheap Compute, and What Happens If Intelligence Gets CommoditizedJun 29
Cline $9.99/mo GLM-5.2 Plan: Bundled Open-Weights Access ExplainedJun 29
DeepSeek V4 Official Release Mid-July 2026: Peak-Hour Pricing ExplainedJun 29
GLM-5.3: Zhipu AI Asks the Community — Vision Leads the WishlistJun 29
Qwen 3.6 27B Local Dev Guide: llama.cpp, OpenCode, and Why Dense Beats MoEJun 29
Top Chinese AI Companies and Startups in 2026: The Complete Landscape GuideJun 29
US vs Chinese AI Startups in 2026: Funding, Strategy, and Who Wins WhatJun 28
Chinese Firms Predict the AI Bubble Is Close to Bursting — What the Warning Signals MeanJun 28
Zhipu AI's New Model Matches Claude Mythos on Security Bug Detection — What It Means for the AI RaceJun 27
DeepSeek DSpark: speculative decoding for V4 Flash and Pro (51–400% faster inference guide 2026)Jun 27
Meta Llama 4: The Complete Open-Source AI Model Guide 2026Jun 26
CoffeeBench: Sakana AI Benchmarks 90-Day LLM Supply Chain ManagementJun 20
Sarvam AI: Full Capabilities Guide — Models, API, Speech, Vision & How to Run (2026)Jun 19
GLM-5.2 vs Claude Fable 5: Kilo Code's Planning Benchmark Shows a Near-Tie at 1/10th the PriceJun 19
How to Run GLM 5.2 in Claude Code, Pi, OpenCode & Every Harness (2026)Jun 16
VibeThinker 3B: A 3-Billion Parameter Model That Matches Opus 4.5 PerformanceJun 15
GLM-5.2 Beats Fable 5 on Reasoning — 24 Hours After the U.S. Export BanJun 10
Cohere North Mini Code: Open-Source Agentic Coding Model (Apache 2.0)
May 2026
May 24
DeepSeek V4 Pro Shakes the AI Industry: 34x Cheaper Than GPT-5.5 and What It Means for 2026May 23
DeepSeek V4-Pro locks in 75% permanent API discount: $0.435/M tokens, 20x cheaper than GPT-5.5May 22
Cohere Command A+: the first fully Apache 2.0 enterprise AI model that runs on 2 H100s (May 2026)May 22
Qwen 3.7-Max: The Agent Frontier and Long-Horizon AutonomyMay 6
DeepSeek-TUI: terminal coding agent for DeepSeek V4 (Rust, MCP, skills)May 4
DeepSeek V4-Pro: agent coding benchmarks, 1M context, and API economics
April 2026
Apr 27
DeepSeek V4 preview: V4-Pro, V4-Flash, 1M context API (2026)Apr 22
What are parameters in a large language model? Billions, MoE, and what 2026 model cards really sayApr 17
GLM-5.1 on Hugging Face & how to run it (Z.ai API, Ollama, vLLM) — 2026 guideApr 17
Netflix VOID on Hugging Face: video object removal that respects physics (model card recap)