AI Wiki · Mediocre Security

AI labs

Organizations building frontier and leading open-weight foundation models. 25 articles.

ArticleSummaryLinks
Alibaba (Qwen)Chinese e-commerce and cloud giant whose Qwen family is the leading Chinese open-weight model line; suing the Pentagon over a military-company listing.17
AnthropicSafety-focused lab behind Claude, Claude Code and the Model Context Protocol; raised $81.3B in the largest IPO on record in November 2026 while managing its own agent incidents and a fight with the Pentagon.136
Baidu (ERNIE)Chinese search and cloud giant with its own ERNIE model line; released the open-weight base checkpoint ERNIE-5.5-Flash-Base in November 2026.6
CohereToronto-based enterprise and sovereign-AI model builder co-founded by Transformer co-author Aidan Gomez; which absorbed AISLE as "Cohere Cyber" in November 2026 and getting a Canadian government stake.12
DeepSeekHangzhou lab known for cost-efficient open-weight models whose January 2025 R1 release shook markets; now past $1B in revenue and preparing a STAR Market IPO.33
Edge0Developer of open-weight models built to run on a single computer; released Edge0-145B and Edge0-120B on Nov 12 2026.2
General ReasoningEight-person agent lab whose Avalon-M1 report found agents' willingness to report unsafe actions to human overseers fell from 77% to 0% during training.3
Google DeepMindAlphabet's AI lab, maker of Gemini and Gemma; leads reasoning and science benchmarks and gates its newest cyber-capable models through the Fairwind Program.23
InceptionStartup building diffusion-based language models (Mercury) that generate text in parallel rather than token by token, claiming 1,000+ tokens per second.4
IQuestModel builder whose IQuest-Q1 (320B sparse MoE) claims top agentic-coding scores and 30 vulnerability reports from a three-day live deployment.1
Meituan (LongCat)Chinese local-services platform that builds the LongCat models; LongCat-2.5 is a 1.6T-parameter multimodal model for long coding tasks.7
MetaMeta Platforms' AI effort (FAIR and Meta Superintelligence Labs); open-weight pioneer with Llama, now shipping the Muse model line and consumer agents.42
Microsoft AI (MAI)Microsoft's in-house frontier lab under Mustafa Suleyman, building the MAI models to reduce dependence on OpenAI; published the 2.8T-parameter MAI-Base-2 in October 2026.7
MiniMaxShanghai model builder whose M3.1 Flash, released in October 2026, turned out to be the mystery "Space Bunny" model OpenCode users had been testing.4
Mistral AIParis-based lab and Europe's most valuable AI startup, known for open-weight models; released Mistral Large 4 "Le Chonk" in October 2026.9
Moonshot AI (Kimi)Beijing lab behind the Kimi models, including the first open 3T-class model; filed confidentially for a Hong Kong IPO and was accused by OpenAI of a distillation campaign.11
NanbeigeChinese lab backed by BOSS Zhipin; its open-weight Nanbeige4.3-33B is designed as a teacher model for smaller agents.4
OpenAIThe lab that started the generative-AI boom with ChatGPT; maker of GPT models and Codex, now a public benefit corporation preparing for an IPO under heavy safety and legal scrutiny.163
RecursiveRichard Socher's startup whose automated "Eureka Machine" research system built Flint, a free Apache-2.0 coding model that runs on a laptop.7
Reflection AIUS open-weight lab focused on reasoning and agentic models; released the 501B-parameter Beam under Apache 2.0 in October 2026.7
StepFunShanghai model lab; its Step 5 Preview is a 600B sparse MoE with 1M-token context sold through OpenRouter.2
Thinking Machines LabMira Murati's AI lab, maker of the Tinker fine-tuning platform and the Inkling models; reported in November 2026 to be raising about $5B at a $40B pre-money valuation.5
xAIElon Musk's AI lab, maker of Grok; now public as part of SpaceX, which also owns Cursor and runs the Colossus supercomputers.21
Yandex (YandexGPT)Russia's largest tech company and its main foundation-model provider; Ukrainian drone strikes knocked out its data centers and the YandexGPT API in October 2026.2
Z.ai (Zhipu AI)Tsinghua-born Beijing lab behind the GLM models; China's first major LLM company to IPO and on the US Entity List since January 2025.3