Google is where large-scale research, Gemini-era multimodality, and public product visibility now converge in the atlas.

OpenAI is where much of the public story around GPT systems became legible through reports, releases, and recognizable research teams.

Moonshot AI matters because Kimi made high-end long-context and multimodal systems feel consumer-facing and publicly discussable.

ByteDance Seed is ByteDance's frontier-model team for reasoning and multimodal systems.

DeepSeek stands out because the public has become deeply curious about how a lean team ships frontier-level reasoning systems.

Qwen is one of the clearest examples of a large industrial lab turning model families into a long-running public research program.

Z.ai represents another branch of the Chinese frontier-model ecosystem, where GLM-style systems carry a distinct public identity.

MiniMax signals how quickly a newer lab can join the frontier conversation with its own model family and public presence.

Anthropic matters because Claude made safety, system behavior, and model governance part of mainstream AI conversation.

NVIDIA matters as an extended layer where systems, infrastructure, and model work intersect in public-facing ways.

Ai2 gives the atlas a different angle: open research, public tooling, and researchers who are often easier to understand through ideas than products.

Apple appears here as an extended frontier actor where reports and product systems shape public understanding differently from pure labs.

Cohere sits in the extended layer as a company with meaningful public work around enterprise LLM systems and open research initiatives.

Mistral is part of the extended frontier group where open-weight releases make researchers more publicly traceable.

Meta AI became especially visible through the open Llama line, where releases also turned into a public map of the people behind them.

Microsoft remains important as an extended layer where research, productization, and platform distribution meet.

Amazon appears here as an extended actor where model development, infrastructure, and product surfaces reinforce each other.

Shanghai AI Laboratory belongs in the extended layer where open technical reports from the InternLM and InternVL lines make a large but traceable researcher set visible.

Baichuan belongs in the extended layer where frontier language-model releases and public technical reports make researchers and report lines legible.

Salesforce AI Research belongs in the extended layer where open multimodal and agentic model reports expose a sizable but still traceable researcher set.

01.AI belongs in the extended layer where publicly released foundation-model reports make a distinct researcher set visible.

Cerebras Systems belongs in the extended layer where hardware-linked model reports still surface identifiable public researchers.

Technology Innovation Institute belongs in the extended layer where open Falcon model reports make a modest but public researcher set visible.

Databricks belongs in the extended layer where open foundation-model releases and public technical reports make a broad but still traceable researcher set visible.

IBM Research sits in the extended layer where long-running industrial research programs still publish public model work with identifiable technical contributors.

Liquid AI belongs in the extended layer where public model-family reports expose a compact but meaningful researcher set.

Snowflake belongs in the extended layer where public model releases and technical reports expose a compact but identifiable researcher set.

BIGAI belongs in the extended layer where open multimodal model releases and public technical reports expose a meaningful research cohort.

Stepfun belongs in the extended layer where public multimodal and speech-model reports make a compact but visible researcher set traceable.

Tencent Hunyuan belongs to the extended layer where major platform companies continue to shape multimodal and large-model work.

AI21 Labs remains part of the extended layer for its role in public-facing language-model development and product experimentation.

Kyutai belongs in the extended layer where open speech and multimodal model releases expose a focused but meaningful research cohort.

LG AI Research belongs in the extended layer where public foundation-model reports make a smaller but identifiable researcher set visible.

OSU NLP Group belongs in the extended layer for UGROUND-style GUI-agent grounding work with public reports and identifiable researchers.

Baidu belongs in the extended layer where major public model launches, technical reports, and developer-facing releases make researchers more traceable.