CATEGORY · 132 COMPANIES
Foundation Models / LLM Providers
Labs training frontier large language and multimodal models accessed via apps and APIs.
#CompanyValuationARRHeadcountFresh
01OOpenAIFrontier AI lab and maker of the GPT and o-series models, ChatGPT, DALL-E, and Sora, accessed via consumer apps and the OpenAI API platform.—~$25B run-rate (early 2026); 2024 revenue $13.1B~3,500+—02AAnthropicSafety-focused frontier lab building the Claude family of models and Claude Code, offered via Claude.ai, the Anthropic API, Amazon Bedrock, and Google Vertex.—$14B+ ARR (early 2026), up from ~$1B 14 months prior~2,000+—03xxAI (Grok)xAI is an artificial intelligence company founded in March 2023 by Elon Musk that develops the Grok family of large language models, an API platform, and consumer/enterprise AI products. The company operates the Colossus GPU (Graphics Processing Unit) supercomputer cluster in Memphis, Tennessee, and in March 2025 acquired the social network X in an all-stock transaction. In February 2026, xAI was acquired by SpaceX, combining the two Musk-led entities.$230B (reported, January 2026 Series E)~$500M Annual Recurring Revenue (ARR) standalone exit-2025 run rate (reported)~1,200–4,900 (reported, varies by source; PitchBook lists 4,900)green04GGoogle DeepMind (Gemini)Alphabet's AI division producing the Gemini model family plus Veo (video), Imagen (image), and Lyria (music), integrated across Google products.—Not separately broken outNot separately reported—05MMeta AI (Llama)Meta's open-weights Llama models and the Meta AI assistant integrated across Facebook, Instagram, WhatsApp, and Ray-Ban Meta glasses.—Not separately broken out; ~$70-80B AI capex 2025Not separately reported—06MMistral AIEuropean open-weights LLM lab producing Mistral Large, Medium, Magistral (reasoning), Devstral (coding), and the Le Chat consumer app.—~$100M ARR (2025), up ~25x YoY~276—07DDeepSeekOpen-weights reasoning LLMs at low cost————08001.AI (Yi)Open Yi model family————09GGeneral IntuitionGeneral Intuition is a frontier AI research lab spun out from Medal (a gaming clip platform) that trains spatial-temporal reasoning models and world models using 2 billion gaming videos annually to enable AI agents to understand how objects move through space and interact with physical environments.———green10EEpistemic AIEpistemic AI applies large language models and AI to biotech, pharmaceutical, and life sciences research. Founded by Stefano Pacifico, the company leverages AI to accelerate drug discovery, molecular modeling, and biomedical research workflows, addressing the post-COVID biotech funding decline.———green11WWaabiWaabi develops Waabi World, a physics-based simulator and autonomous driving system for Level 4 autonomous trucking, with partnerships for real-world deployment and integration into commercial truck fleets.estimated $500M+—~150green12BByteDance DoubaoChina consumer LLM————13IInflection AIAI studio creating emotionally intelligent personal AI assistant Pi. Develops large language models focused on empathetic, conversational interactions for personal and enterprise use across multiple devices and platforms.$4B~$1.8M (2024)63-78green14BBaiduBaidu is a leading AI company and search engine with deep expertise in artificial intelligence, deep learning, natural language processing, and computer vision. The company develops ERNIE large models (ERNIE 4.5, ERNIE X1) deployed at scale across search, information feeds, and smart speaker products, and empowers industries through Baidu AI Cloud.———green15LLatent Labs<cite index="53-17">Latent Labs is an AI-driven biotechnology company developing AI foundation models to make biology programmable. The company innovates on and applies generative AI technologies to enable biotechnology and pharmaceutical companies to computationally create new molecules for therapeutic, industrial, and environmental uses.</cite>——21green16EExaExa is an AI-native search engine that retrieves real-time data from the web optimized for artificial intelligence applications. Unlike traditional search engines built for human users, Exa trains embedding models to help AI systems find the highest-quality knowledge without SEO spam or ads.$2.2B——green17MMoonshot AI (Kimi)Long-context Chinese LLMs————18IImbueReasoning agents research————19AAlibaba QwenOpen LLMs with leading enterprise token share————20GGuide LabsGuide Labs develops interpretable foundation models and AI agents that can explain their reasoning and are easier to align with human intent. The team comprises interpretability researchers responsible for major advances in the field, working to rethink how machine learning models and AI systems are developed.——9green21WWorld LabsLarge World Models for spatial AI————22PPrime IntellectGPU marketplace and distributed training infrastructure startup enabling decentralized AI model training. Trained INTELLECT-1 (10B) and INTELLECT-3 (106B) on distributed GPUs across multiple countries.———green23AAI21 LabsJamba long-context LLMs and enterprise APIs————24SSakana AIEvolutionary model-merging research————25NNous ResearchOpen community model development————26LLiquid AIEfficient edge LLMs (MIT spinoff)————27AAleph AlphaEuropean enterprise AI operating system————28NNebius.AINebius.AI is a European cloud AI platform providing inference APIs, fine-tuning, and model hosting for open-source LLMs and image models. The company emphasizes data privacy, EU data residency, and compliance, targeting organizations requiring GDPR and sovereignty guarantees.—estimated low millions run-rateestimated 10–15green29UUnreal AI (formerly LastMile AI)AI developer platform enabling engineering teams to efficiently create generative AI applications for language, image, and audio models. Unifies tools, promotes collaboration via notebook-like environment for building LLM apps.———green30AAxiom MathAI mathematician and theorem prover developing verified AI systems for autonomous mathematical discovery and formal proof verification using the Lean proof assistant. Raised $200M Series A (March 2026) at $1.6B valuation after $64M seed (October 2025). Achieved perfect score on Putnam 2025 competition.$1.6B—30+green31NNeuralk-AIFrench AI startup pioneering tabular foundation models for structured data processing. Neuralk-AI specializes in applying generative AI to enterprise data warehouses and e-commerce catalogs, addressing the gap where large language models struggle with tabular data.———green32MModularModular develops the Mojo programming language and MAX platform for high-performance AI inference, enabling developers to write GPU kernels in Python-compatible syntax and deploy across NVIDIA, AMD, and other hardware.estimated $1B+—~150green33AArcee AIArcee AI develops secure, enterprise-focused platforms for building, fine-tuning, and deploying custom language models. The company specializes in small language models (SLMs) and recently released Trinity, a frontier open-weight foundation model family trained on sovereign U.S. infrastructure.——~30green34TTerray TherapeuticsAI-first small-molecule drug discovery platform using machine learning to explore chemical space faster. Founded by Stripe co-founder Patrick Collison and biotech veterans, combining generative chemistry with traditional pharmacology workflows.———green35TTwelve Labs Inc.Twelve Labs builds multimodal video understanding foundation models that map natural language to video content including actions, objects, and background sounds. The startup enables developers to create applications for semantic video search, scene classification, summarization, and chapter generation. Founded by Jae Lee and co-founders from the tech industry, the company focuses on enterprise, media, entertainment, and security use cases.——30,000+ developers on platformgreen36PProject PrometheusProject Prometheus, backed by Jeff Bezos, develops AI optimized for engineering and manufacturing. The company focuses on training models for physical-world applications and industrial automation.———green37SSES AISES AI specializes in AI-discovered high-energy, high-power density lithium-metal and lithium-ion rechargeable batteries for electric vehicles, drones, urban air mobility, and energy storage. The company uses Molecular Universe, its proprietary AI agent, to accelerate material discovery and develops Avatar, an AI-powered battery health and safety management system.———green38MMultiverse ComputingSpanish AI company developing compressed versions of open-weight models (OpenAI, Meta, DeepSeek, Mistral). Technology reduces model size and computational cost while maintaining performance.———green39LLuma AIFrontier AI lab building multimodal world models capable of generating, understanding, and operating in physical environments. Develops Dream Machine (30M+ users), Ray3 (reasoning video model with HDR), and Genie (3D generation) for creative professionals and enterprises in entertainment, advertising, and robotics.$4B+~$15-20M (2025 estimate)~500 (expanding to 700+ with London hub)green40DDefog.aiDefog provides specialized Large Language Models (LLMs) for data analysis, including SQLCoder, an open-source model that converts natural language questions into SQL queries. The company's platform enables enterprises to query structured data and data warehouses using natural language while maintaining data privacy through on-premise deployment.—~$100K-5M estimated1-25green41AAxiomAxiom builds an AI reasoning engine for autonomous mathematical discovery, theorem proving, and formal verification using the Lean proof assistant. The company develops verified AI systems that generate formally proven code by combining AI with mathematical certainty. Axiom achieved a perfect 120/120 Putnam 2025 competition score—unprecedented for AI.$1.6B—30+green42CContextual AIEnterprise AI platform for building grounded AI agents and RAG (Retrieval-Augmented Generation) applications. Founded by Douwe Kiela and Amanpreet Singh, former Meta AI Research and Hugging Face researchers, the company focuses on context-aware AI that grounds language models in organizational data for mission-critical applications.$150M—51-200green43GGlaiveGlaive develops client-owned AI models optimized for domain-specific tasks including code generation, content creation, and automated customer support. The company provides intelligence solutions for product builders with focus on model ownership and performance, rather than API-dependent services.———green44GGizaWeb3-native AI platform for deploying trust-minimized machine learning on blockchains using zero-knowledge proofs (ZKML). Enables developers to create verifiable AI agents that autonomously execute on-chain strategies with provable, auditable decisions across DeFi protocols.———green45AArc Institute (Evo)Arc Institute is a nonprofit research organization developing Evo, a genomic foundation model family for biological sequence modeling and design. Evo 1 (7B parameters, Nov 2024) and Evo 2 (40B parameters, March 2026) enable prediction and generative design tasks across DNA, RNA, and protein modalities at single-nucleotide resolution.———green46GGoodfireGoodfire is an AI interpretability research lab that uses mechanistic interpretability to understand, debug, and design AI models. The company's Ember platform decodes neural networks to provide programmable access to model internals, enabling users to identify problematic behaviors, extract insights, and improve performance without black-box approaches.$1.25B—estimated 20–30green47BBaidu ERNIEERNIE 4.5 LLMs————48RRedwood ResearchNonprofit AI safety and alignment research organization developing techniques for controlling advanced AI systems and assessing dangerous capabilities. Pioneered the field of AI control, focusing on empirical methods to mitigate risks from misaligned AI agents through techniques like trusted monitoring.—~$10M (2023 grants)11-50green49BBaichuanLLMs with a medical vertical pivot————50PPhysical Intelligence (PI)Physical Intelligence develops foundation models for robotic manipulation and autonomous systems. Models are trained on large-scale robot movement and interaction data to enable general-purpose physical task execution.$2.7B—100+green51OOctoAI (formerly OctoML)AI infrastructure platform for optimizing and deploying generative AI models at scale. Built on Apache TVM compiler framework. Raised $85M Series C (November 2021) led by Tiger Global, bringing total funding to $132M. Rebranded from OctoML to OctoAI (January 2024) as generative AI focus expanded.——140+green52TTencent HunyuanIn-product LLMs across Tencent————53SStability AIOpen-source generative AI company developing Stable Diffusion and multimodal models for text-to-image, video, audio, and 3D generation. Raised $80M Series A extension (June 2024) at $1B valuation. Total funding $181M+ with leadership transition to Prem Akkaraju (ex-Weta Digital CEO, June 2024). 150M+ Stable Diffusion downloads.$1B—184-192green54MMiniMax (Hailuo)<cite index="3-16,3-17,3-18">Founded in 2021 by former SenseTime engineers in Shanghai, MiniMax is valued at $2.5 billion and backed by Alibaba and Tencent. It specializes in text-to-video and conversational AI, rivaling global leaders like OpenAI. A text-to-video AI generator creating high-quality videos for creative industries, competing with OpenAI's Sora.</cite>$2.5B~$70M (2024, via Talkie product)300-500 estimatedgreen55PPredibasePredibase is a low-code platform for fine-tuning and serving open-source language models. The company developed LoRAX framework enabling teams to deploy hundreds of fine-tuned models from a single GPU, dramatically reducing infrastructure costs for LLM customization.——25-35green56AAxiom Quant Inc.Axiom develops verified artificial intelligence systems using formal mathematics and proof assistants (Lean). The company builds autonomous mathematicians for code verification, finance, and chip design, generating formally verified outputs that eliminate hallucinations through machine-checkable proofs.$1.6B——green57LLila SciencesLila Sciences develops AI Science Factories—autonomous laboratories that combine artificial intelligence, software, and robotics to accelerate scientific discovery across life sciences, chemistry, and materials science. The platform enables AI agents to design and run thousands of parallel experiments, compressing workflows that traditionally take weeks into hours.$1.3 billion—~100+green58CCradle BioProtein engineering AI company using generative models to design improved biologics. Enables rapid design of proteins for therapeutics, diagnostics, and industrial biotech applications without extensive manual iteration.estimated $500M—80+green59GGensynDecentralized machine learning compute protocol enabling distributed training with cryptographic verification. Launched Delphi prediction market settled by AI models on Ethereum Layer 2, providing infrastructure for AI model verification and deployment without centralized control.$1B——green60VValence DiscoveryValence Discovery applies AI-driven generative chemistry to accelerate drug discovery for previously intractable biology targets. The platform uses few-shot learning and proprietary molecular representation techniques to enable rapid end-to-end design of novel chemical matter against complex disease targets.———green61CChai DiscoveryAI-native biotech company that predicts and reprograms interactions between biochemical molecules for drug discovery. Builds foundation models tuned to design novel antibodies and therapeutic molecules with improved developability properties, tackling historically hard-to-drug targets.$1.3B——green62IInsilico MedicineClinical-stage biotech using generative AI for end-to-end drug discovery. Proprietary Pharma.AI platform integrates LLMs, multimodal foundation model (Nach01), and Dora research assistant. First fully AI-discovered molecule in Phase 2 trials.——180+green63ZZhipu AI (GLM)LLMs; listed on Hong Kong Stock Exchange————64AAllen Institute for AI (Ai2)Ai2 is a nonprofit AI research institute founded by Paul Allen that emphasizes fully open-source AI development. It releases complete training data, code, and model weights alongside foundation models like OLMo, Tülu (instruction-tuning), and Molmo (multimodal vision-language), published through 2024–2025.——~300 researchers and staffgreen65AARC Solutions (ARC Reactor)Privacy-first AI infrastructure company developing Reactor, an efficient foundation model trained with 8 NVIDIA L4 GPUs and 4 A100 GPUs. ARC prioritizes data privacy, security, and energy efficiency while delivering competitive performance for enterprise and individual deployment.———green66LLAION<cite index="37-11,37-12">LAION (Large-scale Artificial Intelligence Open Network) was founded in 2021 with the goal of increased access to quality, AI-friendly datasets and curates and releases massive openly-licensed datasets for AI/ML usage.</cite> <cite index="38-2">It is a German non-profit organization.</cite>———green67RReka AIMultimodal foundation models————68HHyperbolic LabsDecentralized GPU inference network enabling cost-efficient model serving and fine-tuning. Provides distributed compute infrastructure for LLM inference, image generation, and video processing by aggregating spare GPU capacity globally, bypassing traditional cloud bottlenecks.Estimated $100-200MEstimated $2-5M (early stage)~30green69HHumans&Human-centric AI startup founded by Anthropic, xAI, and Google researchers. Focuses on AI that empowers people rather than replaces them, with emphasis on reinforcement learning, multi-agent systems, and memory.$4.48B——green70TThinking Machines LabThinking Machines Lab is an AI research organization co-founded by former OpenAI CTO Mira Murati developing broadly capable artificial intelligence systems. The lab focuses on building next-generation AI models with emphasis on reasoning, safety, and real-world applicability.$50-60 billion—~100+green71UUnify AIUnify is a unified API and marketplace for accessing multiple large language models (open and proprietary) through a single endpoint. The platform abstracts model provider complexity, enables model-agnostic applications, and provides fallback routing for reliability.$50M—~15green72DDescrybeDescrybe is a generative AI platform for the legal industry, offering plain-language legal research and a reasoning engine called DescrybeLM. It provides citation-backed legal analysis across 6+ million federal and state court opinions with an AI Cytator tool and brief-checking capabilities launched March 2026.———green73RRecursive SuperintelligenceLondon and San Francisco-based AI research lab developing self-improving artificial intelligence systems that automate the entire AI development pipeline. The company's approach enables autonomous training systems to independently improve their own neural architectures, training algorithms, and evaluation protocols.$4.65B—25-30green74AAlignment Research CenterNonprofit research organization focused on AI alignment and safety, developing theoretical methods to ensure advanced AI systems behave in ways aligned with human values. Specializes in evaluations assessing dangerous capabilities and theoretical research on scalable oversight.———green75TTwelve LabsTwelve Labs develops multimodal video understanding APIs that extract contextual information from video content (visual, audio, text). Products include Marengo (multimodal embedding model), Pegasus (search/classification), and Embed API for developers. Integrated into Amazon Bedrock and Databricks.———green76MMind RoboticsPalo Alto-based robotics company spun out by Rivian co-founder RJ Scaringe, developing a full-stack platform combining foundation models, purpose-built robotics hardware, and deployment infrastructure for industrial manufacturing automation. The company partners with Rivian to deploy AI-powered robots in real production environments.$2B—100+green77IIneffable IntelligenceLondon-based AI frontier lab founded by former Google DeepMind reinforcement learning lead David Silver, pursuing superintelligence through reinforcement learning systems that discover knowledge from experience without relying on human data. The company aims to create a 'superlearner' capable of continuous self-improvement.$5.1B—50+green78ZZ.aiZ.ai (formerly Zhipu AI) is a Beijing-based AI company founded by Tsinghua University researchers that develops the GLM (General Language Model) family of open-weight models optimized for multimodal reasoning, coding, and agentic execution. The company has shifted from foundation model focus to building comprehensive application layers including ChatGLM (13M+ monthly users), Ying video generation, and AutoGLM autonomous agents for smart devices.$20B+~$44.5M (2024)250-500 estimatedgreen79GGradiumGradium is developing next-generation expressive artificial intelligence voice models capable of conveying emotion, nuance, and personality in speech synthesis. The company focuses on building voice models that sound more natural and human-like than existing text-to-speech technologies.———green80SSambaverse / SambaNova SystemsSambaNova Systems develops AI accelerator hardware and software for large-scale foundation model training and inference. The company offers specialized chips and cloud services optimized for generative AI workloads.~$5.2B—~400green81HHarmonic AIHarmonic AI develops formal verification and mathematical reasoning AI systems, competing in the verified AI space alongside Axiom. Focuses on theorem proving, formal mathematics, and safety-critical AI verification.$1.45B——green82CCohere (post-Aleph Alpha acquisition)Cohere is an enterprise AI lab developing foundation models and custom solutions for regulated industries. After acquiring German AI company Aleph Alpha (April 2026), Cohere expanded its sovereign AI and European public-sector positioning with dual headquarters.$7B~$100MNot separately reportedgreen83AAdvanced Machine Intelligence (AMI)AI research lab founded by Yann LeCun (former Meta chief AI scientist) developing world models and reasoning-based AI systems designed to learn structured representations of real-world environments and plan behavior sequences for robotics and complex applications.$3.5B—small teamgreen84SSarvam AISarvam AI develops foundational large language models (LLMs) and multimodal AI systems specifically trained on Indian languages and region-specific use cases. The company released Sarvam-30B and Sarvam-105B models in February 2026 and announced Sarvam Kaze, an indigenous AI-powered smart glass wearable, launching in May 2026.$1.5B–$1.55B~$3.6M (FY2024-25)~200–250green85AAdept AIAdept develops multimodal AI agents that automate digital workflows, integrating AI models with custom actuation software to execute repetitive enterprise tasks. Models include ACT-1 (for text commands) and ACT-2 (for visual manipulation of software interfaces).$1B~$75M69green86RReplisomeReplisome develops context-efficient open-source foundation models and training methodologies. The company focuses on reducing computational overhead and making large-scale model training accessible to resource-constrained teams through novel optimization techniques.——~20green87PPeriodic LabsAI scientist platform focused on autonomous research, hypothesis testing, and scientific discovery. Aims to create agentic AI systems capable of conducting independent research across scientific domains.——20-50green88TTARS RoboticsTARS Robotics is a Chinese humanoid robotics startup developing general-purpose humanoid robots for industrial and commercial applications. The company focuses on embodied AI and robot manipulation.———green89SSandboxAQAI research lab and quantum-classical hybrid model company developing AI for quantum advantage. Raised $450M Series E at $5.7B valuation with backing from Nvidia, Google, and Ray Dalio.$5.7 billion—300+ estimatedgreen90VVlex/Vincent AIvLex (now part of Clio) is a comprehensive legal intelligence platform with Vincent AI, a precision-engineered AI for legal research, litigation, and transactions. It accesses over 1 billion legal documents from 100+ countries with litigation analytics covering 850+ million court records and advanced search across multi-jurisdictional legal content.———green91RRunway ML<cite index="8-26,8-27">Runway ML, Inc. is an American company headquartered in New York City that specializes in generative artificial intelligence research and technologies, primarily focused on creating products for generating videos, images, and various multimedia content through developing commercial text-to-video and video generative AI models.</cite>$5.3B~$300M projected 2025~140green92CClarifaiClarifai is an end-to-end AI platform specializing in computer vision, natural language processing, and audio recognition. It provides tools for data labeling, model training, evaluation, and inference, supporting the full AI lifecycle with pre-trained models and custom model development for image, video, text, and audio data.———green93EEmmi AIEmmi AI merges physics and artificial intelligence to accelerate industrial engineering workflows, replacing slow traditional simulations with intelligent AI-powered tools. The company develops models for computational fluid dynamics and aims to expand into multi-physics real-time simulations for aviation, energy, and automotive sectors.——undisclosedgreen94TTencent Hunyuan VideoHunyuan Video is Tencent's text-to-video and image-to-video generative AI model that delivers studio-grade output with high motion fidelity and scene consistency. <cite index="11-16,11-17">Hunyuan Video delivers studio-grade text-to-video and image-to-video results with high motion fidelity and scene consistency. The model excels at generating continuous, director-style sequences from a single prompt, so you get complete actions and native camera cuts instead of disjointed clips.</cite>———green95CCuspAICuspAI is a Cambridge-based AI startup developing a platform for next-generation materials using generative AI, deep learning, and molecular simulation to accelerate material design. The company uses AI to invert the traditional materials discovery process, starting with desired outcomes rather than trial-and-error experimentation.———green96iiFlytek<cite index="37-24,37-25">iFlytek is a partially state-owned Chinese information technology company established in 1999. It creates voice recognition software and 10+ voice-based internet/mobile products covering education, communication, music, intelligent toys industries.</cite> The company has pivoted into large language models (iFlytek Spark/Xinghuo) and multimodal agents, with widespread deployment in healthcare diagnostics, education, and smart devices.Market cap varies; substantial public enterpriseUndisclosed; multi-billion yuan revenue at enterprise scale3,000+green97AArgillaArgilla is an open-source data labeling platform for Natural Language Processing that empowers data and machine learning teams to build and monitor high-quality training data. The platform provides human-in-the-loop and programmatic labeling features, and was acquired by Hugging Face in June 2024.———green98DDecartAI research lab developing advanced foundation models and reasoning systems. Focuses on efficient model architectures and novel training approaches competitive with major foundation model labs.$3.1B——green99PPatronus AIPatronus AI provides AI evaluation and guardrailing solutions for large language models. The company's platform helps enterprises test, benchmark, and deploy LLMs safely by detecting hallucinations, evaluating factuality, and automating quality assurance across AI applications.——estimated 15–20green100PPhi-4 (Microsoft Research)Dense 14-billion parameter language model developed by Microsoft Research emphasizing high-quality training data and advanced reasoning capabilities. Released December 2024 under MIT license for efficient on-device AI.———green101AAxiomatic AIAI mathematics verification startup focused on formal verification of AI-generated proofs and reasoning. Secured $18M for math verification efforts, competing in similar space to Axiom Math.———green102NNetEase YoudaoNetEase Youdao is an intelligent learning company and subsidiary of NetEase focused on large language models for education. It developed Ziyue, a vertical foundation model for K-12 and adult learning, with specialized reasoning variants (Ziyue-o1) released in 2025. The company deploys models across smart hardware, translation, tutoring agents, and content generation.~$1–2B implied (subsidiary of NTES; not independently valued)~$224M (2024 reported revenue)Not disclosedgreen103SSafeAI (Formerly Flo Health / AI Safety-Focused)Specialized AI safety and alignment research lab founded by industry veterans with focus on creating safe, interpretable foundation models and addressing critical AI risks.———green104SStepFunShanghai-based multimodal AI startup founded by former Microsoft employees. Developed Step-2, a trillion-parameter LLM with vision and image generation (Step-1X) capabilities. Positioned as one of China's 'AI Tiger' companies.——~97green105MMistral VoxtralVoxtral is a family of open-weight audio AI models developed by Paris-based Mistral AI, encompassing speech-to-text (ASR = Automatic Speech Recognition), speech understanding, and text-to-speech (TTS) capabilities. First released in July 2025 as Mistral's debut audio model line, the family has expanded to include Voxtral Transcribe 2 (February 2026) and Voxtral TTS (March 2026), positioned as lower-cost open alternatives to closed providers such as OpenAI Whisper and ElevenLabs.$13.8B~$100M-$200M (Mistral AI parent, reported)~500-1,093 (Mistral AI parent)green106MModelBestA Tsinghua University-affiliated spinoff commercializing the MiniCPM family of efficient on-device foundation models. MiniCPM models are engineered for smartphones, PCs, automotive systems, and edge devices with GPT-4V-level multimodal performance at 1-9B parameters. Operates OpenBMB, an open-source research lab for large model development.———green107SSenseTimeChina's leading computer vision company founded by Tang Xiao'ou and Li Xu, SenseTime pivoted decisively to generative AI. In 2025, generative AI revenue surged 51% to 3.63B yuan (~$527M), comprising 70%+ of total revenue. The company achieved first positive EBITDA post-IPO in H2 2025 with SenseNova multimodal LLM and enterprise productivity tools (Raccoon series).~$15B+ (public: 0020.HK)~$728M (2025)~1,887green108AAI2 IncubatorAI2 Incubator is an early-stage startup incubator and venture arm of the Allen Institute for AI (Ai2), a nonprofit research institute founded by Microsoft co-founder Paul Allen. The incubator co-founds and mentors AI-first companies, providing capital, cloud credits, and access to frontier AI research.——~10 managing directors/technical staff per public sourcesgreen109SSafe Superintelligence Inc. (SSI)Safety-focused frontier AI research lab co-founded by former OpenAI chief scientist Ilya Sutskever. Singular mission: develop safe superintelligent AI systems with all resources aligned to AGI research and safety, avoiding premature commercialization.$32.0B—~50green110GG42Emirati artificial intelligence development holding company focused on AI development across government, healthcare, finance, oil and gas, aviation, and hospitality. Chaired by UAE National Security Advisor Tahnoon bin Zayed Al Nahyan.——~800+green111IIntellaArabic speech-recognition and analytics platform specializing in dialect transcription (Egyptian, Gulf, Levantine). Trains models on regional audio from call centres and media archives. Offers real-time transcription, sentiment/intent analysis, voicebots, and agent-assist via APIs.———green112BBotlhale AISpeech and language model builder for Southern Africa specializing in understanding code-switching, slang, and colloquial speech. Offers contact-centre AI stack (transcription, intent detection, routing, analytics) via APIs for banking/telecom.———green113HH Company<cite index="11-2,11-3">H Company was founded in 2023 in Paris by Laurent Sifre, Charles Kantor, and three DeepMind veterans: Daan Wiestra, Karl Tuyls, Julien Perollat.</cite> <cite index="11-9">H Company specifically designs AI agents called computer use capable of autonomously interfacing with any software (local or cloud-based) to detect and automate repetitive operations.</cite> <cite index="11-14,11-15">In March 2026, H Company released Holo3, a family of artificial intelligence models designed to operate digital systems by interacting directly with user interfaces.</cite>——~330green114LLightOn<cite index="36-15,36-16,36-17">LightOn SA is a French company focused on artificial intelligence that develops generative AI solutions for enterprises. Its current product is the Paradigm platform, an on-prem/secure Enterprise generative AI platform. In November 2024, it became Europe's first listed GenAI startup to launched an IPO on Euronext Growth Paris.</cite>—<cite index="31-21">€1.13M annual revenue as of Dec 31, 2024</cite>—green115LLinagora<cite index="49-1,49-2">Linagora is a French open source software editor, founded in June 2000 by Alexandre Zapolsky and Michel-Marie Maudet. Located in France, as well as in Belgium, Canada, Vietnam, the United States and Tunisia, the company employs around 200 people.</cite> <cite index="49-4,49-5,49-6">In July 2023, Linagora launched OpenLLM France, a community comprising many actors focused on generative AI, to develop a sovereign and open source large language model. This initiative, led by co-founder and CEO Michel-Marie Maudet, had more than four hundred French members by early 2024.</cite>——~200 employeesgreen116AAllganizeKorean AI company developing industry-specific large language models (e.g., finance, legal, manufacturing). KPAS 2024 finalist. Part of government sovereign AI push to reduce reliance on foreign models.——30-60green117HHuawei Pangu (Huawei Cloud Pangu Models)<cite index="19-19,19-20">Huawei PanGu is a multimodal large language model developed by Huawei and officially launched in July 2021.</cite> <cite index="23-1,23-2">Huawei Cloud released Pangu Models 5.5, a 718-billion parameter AI platform designed specifically for industrial applications across manufacturing, agriculture, and scientific research, with applications that have delivered measurable productivity gains across more than 500 scenarios in over 30 industries.</cite> <cite index="26-25,26-26,26-28">The Pangu models are designed for ToB markets and feature a three-layer architecture—L0 comprises five foundation models, L1 includes industry-tailored models, and L2 targets scenario-specific models.</cite>———green118BBioptimusBioptimus builds universal foundation AI models for biology, integrating data across molecular to organismal scales to accelerate drug discovery and precision medicine. The company released H-Optimus-0, a foundational model for pathology trained on over 500,000 histopathology slides, with plans to launch M-Optimus, a multi-scale, multi-modal model simulating complex biological processes.undisclosed (Series A+ 2024)—undisclosedgreen119KKrutrimKrutrim, founded by Ola founder Bhavish Aggarwal, was India's first AI unicorn. Initially positioning itself as a full-stack sovereign AI player with multilingual LLMs, the company pivoted in late 2025 to focus on AI cloud infrastructure services for Indian enterprises while pausing chip and foundation model development.$1B (Series A)~$37.5M (FY2025-26)300–400 (estimated)green120PPhysicsX<cite index="9-1,9-13">Industrial AI-startup PhysicsX has raised an impressive £99.9 million in Seed B financing, bringing its total funding to almost £125 million. Founded by Robin Tuluie, the former head of R&D at Renault F1 and Mercedes F1, and Jacomo Corbo, Vehicle Technology Director at Bentley Motors, the London-based company aims to streamline the engineering and manufacturing industries with AI.</cite>$1B—~150green121AAdaptive MLAdaptive ML develops an LLM platform enabling continuous improvement of language models by learning from real-time user interactions. The platform allows businesses to personalize AI outputs, enhance decision-making efficiency, and train/deploy language models using user feedback at scale.———green122AAlibaba (Damo Academy)Alibaba's research lab (Data, Algorithm, and Machine-learning Optimization) conducts foundational AI research including language models, vision models, and novel architectures. Operates parallel to Qwen team; publishes peer-reviewed research and contributes to proprietary model development.—Research-focused; integrated into Alibaba parentNot disclosedgreen123HHuawei Cloud / Pangu ModelsHuawei's cloud division operates Pangu, a family of multimodal industrial AI models designed for manufacturing, finance, and transportation. Pangu 5.5 (718B parameters) deployed across 500+ scenarios in 30+ industries with production-grade inference infrastructure.—~$10B+ (Huawei Cloud segment, 2024)2000+green124UUpstageUpstage is a leading Korean AI foundation model developer competing in South Korea's sovereign AI initiative. <cite index="13-1">The company was selected by the Ministry of Science and ICT as one of five consortia to compete for sovereign AI development, with $381 million in government funding allocated</cite>.———green125NNyonicFoundation models for European industry with focus on multilingual, domain-specific LLMs for manufacturing and automotive. Develops vertical models tailored to mechanical engineering and industrial use cases. Emphasizes trustworthy, sovereign AI aligned with EU regulations.$100-150M (estimated)—~50green126AAMI Labs (Advanced Machine Intelligence)AMI Labs, founded by Turing Award winner Yann LeCun, focuses on building AI systems that understand the physical world through world models. The company emphasizes 3D understanding and reasoning capabilities beyond token prediction, enabling robots and physical AI systems to comprehend causality, physics, and spatial relationships.———green127MMila – Quebec Artificial Intelligence InstituteWorld's largest academic AI research center specializing in deep learning and machine learning, with 1,500+ members across multiple Quebec institutions. Now actively commercializing research through Mila Ventures and the Venture Scientist Fund to bridge research-to-market gap.——~1,500+ researchers and facultygreen128RRetrievaRetrieva develops Japanese-focused natural language processing AI systems for research, information organization, and analysis. Products include Yoshina (text analysis) and Tsunade (internal document search). The company provides speech recognition, bioinformatics, and call center support solutions.——11-50green129CConverge BioGenerative AI platform accelerating drug discovery and development by automating target identification, antibody design, and protein optimization. Works across oncology, autoimmune, and neurodegenerative research.——34 employeesgreen130LLelapa AIAI language model builder focused on African languages including Swahili, Yoruba, IsiXhosa, Hausa, IsiZulu with 350M+ speakers. Develops region-specific models for greater linguistic accuracy and cultural relevance.———green131AAqemia<cite index="1-6,1-7">Aqemia combines generative AI with deep physics simulations to revolutionize drug discovery at the molecular level. Founded in 2019, it has raised over $100 million and, in 2025, expanded its operations to London as it moves closer to launching clinical trials for its AI-discovered drug candidates.</cite> <cite index="1-5">Reduces the energy consumption of AI training by up to 80%, enabling scalable and eco-friendly AI deployments.</cite>———green132EEvolutionaryScale (ESM3)EvolutionaryScale is a frontier AI research lab for biology, founded by members of the team that built Meta's ESM protein language models. Its flagship model ESM3 is a generative model that reasons jointly over protein sequence, structure, and function to design novel proteins.———green