Together AI

Together AI is an American artificial intelligence cloud and infrastructure company founded in June 2022 by Vipul Ved Prakash, Chris Ré (Stanford computer-science professor), Ce Zhang (then ETH Zürich, now University of Chicago), Percy Liang (Stanford and director of CRFM), and Tri Dao (Princeton CS professor and creator of FlashAttention). The company is headquartered in San Francisco and operates an AI-native cloud platform providing GPU compute, foundation-model training, fine-tuning, and inference services to AI developers and enterprise customers. Together AI is one of the principal commercial alternatives to AWS, Azure, and Google Cloud for AI infrastructure, and has produced open-source research including RedPajama (the open replication of the LLaMA training dataset) and Striped Hyena (the hybrid attention-and-state-space-model architecture). As of February 2025, Together AI has raised approximately $534 million across multiple funding rounds at a $3.3 billion valuation.

At a glance

• Founded: June 2022 in San Francisco by Vipul Ved Prakash, Chris Ré, Ce Zhang, Percy Liang, and Tri Dao.
• Status: Private. Series B in February 2025 valued the company at $3.3 billion.
• Funding: Approximately $534 million cumulative across four reported rounds. Series A of $102.5 million in November 2023 led by Kleiner Perkins with NVIDIA participation. Series B of $305 million in February 2025 at $3.3 billion valuation.
• CEO: Vipul Ved Prakash, Co-Founder and CEO. Technology entrepreneur with previous founding-team participation at Apple Topsy (acquired by Apple in 2013) and other organizations.
• Other notable leadership: Ce Zhang (Co-Founder, Chief Technology Officer; University of Chicago professor), Chris Ré (Co-Founder; Stanford computer-science professor and MacArthur Fellow), Percy Liang (Co-Founder; Stanford CRFM Director), Tri Dao (Co-Founder; Princeton CS professor and FlashAttention creator).
• Open weights: Yes. Together AI's research outputs are released open-weights through Hugging Face under the togethercomputer organization. The company's commercial focus is on AI infrastructure rather than proprietary model development.
• Flagship outputs: RedPajama (open-source LLaMA training-data replication and 3-billion-parameter open-weights model, 2023), Striped Hyena (hybrid Hyena state-space-and-transformer architecture, December 2023), Sequoia (open-weights model, August 2024), BASED (December 2024), Together Inference Engine (the principal commercial inference platform), Together GPU Clusters (the principal commercial training infrastructure).

Origins

Together AI was founded in June 2022 by Vipul Ved Prakash, Chris Ré, Ce Zhang, Percy Liang, and Tri Dao. The founder cohort combined commercial entrepreneurship (Prakash) with deep academic research credentials (Ré at Stanford, Zhang at ETH Zürich, Liang at Stanford, Dao at Princeton). The founding thesis emphasized building open-source foundation-model research alongside the AI cloud infrastructure required to train and deploy such models, with explicit positioning as the open-source alternative to closed-weights frontier-lab approaches.

The 2022 to 2023 period built Together AI's research and infrastructure base. The company assembled GPU compute resources, built the Together training and inference infrastructure, and produced research outputs through collaboration with academic institutions. The May 2023 $20 million seed round, with participation from Lux Capital, Factory, and other investors, provided initial commercial capital.

The April 2023 RedPajama project, a collaboration between Together and several groups including MILA Québec AI Institute, Stanford's CRFM, and ETH's data science lab, was the company's principal open-source contribution. RedPajama replicated the LLaMA training dataset (1.2 trillion tokens of open-licensed text) and trained a 3-billion-parameter open-weights model. The project demonstrated open-source replication of frontier-lab training methodology and contributed to the broader open-source AI ecosystem.

The November 2023 Series A of $102.5 million led by Kleiner Perkins with NVIDIA participation expanded Together AI's infrastructure investment. The December 2023 Striped Hyena release was a structurally distinctive open-source research contribution: a 7-billion-parameter hybrid model combining Hyena state-space-model components with conventional transformer attention components, demonstrating the viability of hybrid architectures at scale.

The 2024 to 2025 period saw Together AI expand commercial operations as one of the principal AI cloud alternatives to the major cloud platforms. The February 2025 Series B of $305 million at $3.3 billion valuation reflected commercial traction in the AI-developer-and-enterprise market. The company's investor base by Series B included Kleiner Perkins, NVIDIA, Emergence Capital, General Catalyst, Salesforce Ventures, Coatue, Lux Capital, Prosperity7 Ventures (the Saudi Aramco-affiliated VC firm), and other organizations.

The 2025 to 2026 period has continued Together AI's commercial expansion alongside continued research output. The company operates GPU clusters across multiple geographies, provides managed inference for over 200 open-source models on the Together platform, and offers fine-tuning and training services for enterprise customers. The Sequoia (August 2024) and BASED (December 2024) research releases continued the company's open-source contribution alongside the commercial AI cloud business.

Mission and strategy

Together AI's stated mission is to build the AI cloud for the open-source AI ecosystem, providing the GPU infrastructure, training capabilities, and inference services that AI developers need without lock-in to proprietary closed-weights model providers. The mission has been remarkably consistent since the founding period.

The strategy combines four threads. First, AI cloud infrastructure including GPU clusters, the Together Inference Engine, and other training-and-inference services. Second, open-source research outputs (RedPajama, Striped Hyena, Sequoia, BASED) demonstrating the company's research capability and contributing to the broader open-source AI ecosystem. Third, the Together platform for hosted inference of over 200 open-source models, providing AI developers with managed access to the leading open-weights models. Fourth, fine-tuning and custom-training services for enterprise customers building specialized AI capabilities.

The competitive premise is that the open-source AI ecosystem requires dedicated AI cloud infrastructure that is structurally distinct from the major cloud platforms (AWS, Azure, Google Cloud) which compete with their own closed-weights AI offerings. Together AI's positioning as the principal open-source-aligned AI cloud provides AI developers with infrastructure that does not compete with their model choices.

The strategic premise has been validated by commercial traction. Together AI is widely characterized in industry coverage as one of the principal AI cloud alternatives to the major cloud platforms, with customer base across AI startups, research organizations, and enterprise customers building with open-source foundation models.

Models and products

Together AI's outputs combine open-source research and commercial AI cloud infrastructure:

• Together GPU Clusters. Reserved GPU compute infrastructure for AI training and inference. Provides developer access to NVIDIA H100, H200, and other GPU configurations.
• Together Inference Engine. Managed inference platform for over 200 open-source AI models including Llama, Mistral, DeepSeek, Qwen, and other open-weights models.
• Together Fine-Tuning. Managed fine-tuning service for customizing open-weights models on enterprise data.
• Together Training. Custom training service for organizations building specialized foundation models.
• RedPajama. Open-source LLaMA training-data replication (1.2 trillion tokens) and 3-billion-parameter open-weights model. Foundational open-source AI research contribution.
• Striped Hyena. December 2023 open-weights hybrid model. Combines Hyena state-space-model components with conventional transformer attention. Demonstrated viability of hybrid architectures at scale.
• Sequoia. August 2024 open-weights research release.
• BASED. December 2024 open-weights research release.
• Together Code Interpreter. AI-augmented code-execution service for AI developer workflows.

The principal commercial channels are direct enterprise sales for Together GPU Clusters, the Together Inference Engine self-service platform, and fine-tuning and custom-training engagement with enterprise customers.

Benchmarks and standing

Together AI's research outputs (RedPajama, Striped Hyena, Sequoia, BASED) are widely cited in academic AI research and have been influential in the broader open-source AI ecosystem. RedPajama in particular validated the open-source-replication thesis for frontier-lab training methodology and provided a foundational training dataset for subsequent open-source models.

Together AI's commercial standing is anchored by the AI developer customer base, the Together Inference Engine's managed-inference scale, and the GPU cluster infrastructure. Industry coverage frequently characterizes Together AI as one of the principal AI cloud alternatives to AWS, Azure, and Google Cloud for AI workloads, alongside CoreWeave (the publicly listed peer), Lambda Labs, and other providers.

The company's standing in the global AI ecosystem is anchored on the founder-team academic-and-commercial credentials, the open-source research-output legacy, and the AI cloud commercial traction.

Leadership

As of April 2026, Together AI's senior leadership includes:

• Vipul Ved Prakash, Co-Founder and Chief Executive Officer. Technology entrepreneur. Previous founding-team participation at Apple Topsy and other organizations. Senior public face for Together AI on company strategy and the AI cloud commercial direction.
• Ce Zhang, Co-Founder and Chief Technology Officer. University of Chicago computer-science professor with machine-learning systems research background.
• Chris Ré, Co-Founder. Stanford computer-science professor and MacArthur Fellow (2015). Senior figure in machine-learning systems research and a serial entrepreneur (DeepDive, SambaNova, and other organizations).
• Percy Liang, Co-Founder. Stanford computer-science professor and Director of the Center for Research on Foundation Models (CRFM) at Stanford HAI. Continues academic appointment alongside Together AI involvement.
• Tri Dao, Co-Founder. Princeton CS professor. Creator of FlashAttention (the memory-efficient attention algorithm that has become a standard component of large-language-model training). Continues academic appointment alongside Together AI involvement.

The company has hired aggressively across AI infrastructure, machine-learning research, and enterprise sales, with recruitment from Stanford, Princeton, ETH Zürich, and other academic and industry organizations.

Funding and backers

Together AI's funding history through April 2026 includes approximately $534 million across multiple rounds. Reported rounds:

• Seed (May 2023): $20 million from Lux Capital, Factory, and other investors.
• Series A (November 2023): $102.5 million led by Kleiner Perkins with NVIDIA participation.
• Follow-on Series A (March 2024): Round bringing valuation to $1.25 billion.
• Series B (February 2025): $305 million at $3.3 billion valuation, with continued participation from Kleiner Perkins, NVIDIA, Emergence Capital, General Catalyst, Coatue, Salesforce Ventures, Lux Capital, Prosperity7 Ventures, and other investors.

The investor base includes NVIDIA strategic-investor participation across multiple rounds, reflecting NVIDIA's broader strategic-investment portfolio across the AI ecosystem. Salesforce Ventures and Coatue strategic-investor participation provides additional capital alignment with the broader enterprise software and growth-equity ecosystems.

Industry position

Together AI occupies a structurally distinctive position in the global AI cloud infrastructure landscape. The combination of the founder-team academic credentials (with three Stanford-and-Princeton professors among the co-founders), the open-source research-output portfolio (RedPajama, Striped Hyena, Sequoia, BASED), the AI cloud commercial scale, and the open-source-aligned positioning produces a profile that no other AI cloud organization matches at the same combination of attributes.

Industry coverage frequently characterizes Together AI as the principal open-source-aligned AI cloud provider, with the company's combination of research credibility and commercial infrastructure scale providing AI developers with an alternative to the major cloud platforms. The strategic positioning has been validated by commercial traction and the $3.3 billion valuation.

Strategic risks include intensifying competition from major cloud platforms (AWS, Azure, Google Cloud) expanding their open-source AI offerings, from peer AI cloud organizations (CoreWeave, Lambda Labs, Replicate, Modal, Fireworks AI), and from frontier-AI labs expanding their inference offerings. Strategic strengths include the founder-team academic-and-commercial credentials, the open-source research-output legacy, the NVIDIA strategic-investor relationship, and the integrated platform value proposition combining infrastructure with research.

Competitive landscape

Together AI competes with several AI infrastructure organizations:

• CoreWeave. Publicly listed AI cloud peer (NASDAQ: CRWV). Direct competitor on GPU infrastructure for AI workloads.
• AWS, Microsoft Azure, Google Cloud, Oracle Cloud. Major cloud platforms competing on AI compute infrastructure. Together AI's open-source-aligned positioning is structurally distinct.
• Lambda Labs, Replicate, Modal, Fireworks AI, Anyscale. AI cloud and infrastructure peers competing in adjacent submarkets.
• NVIDIA Research. Strategic-investor and infrastructure partner. Less direct competition given NVIDIA's broader compute-platform focus versus Together AI's AI cloud focus.
• Hugging Face. Distribution-platform peer with adjacent inference services. Collaboration through Hugging Face's role in distributing open-source models that Together AI hosts.
• Databricks, Snowflake AI, Salesforce AI Research. Enterprise AI platform peers; less direct competition given Together AI's AI cloud infrastructure focus versus the enterprise-application-AI focus.
• OpenAI, Anthropic, Google DeepMind. Frontier-AI labs whose closed-weights APIs are alternatives to the open-weights models hosted on Together AI.

Outlook

Several open questions affect Together AI's trajectory in 2026 and 2027:

• The continued expansion of the AI cloud commercial business, including GPU cluster utilization and inference platform growth.
• The possible Series C round at higher valuation given the market position and continued AI infrastructure demand.
• Continued open-source research outputs and the development of the in-house Together model line.
• The competitive dynamic with CoreWeave (the publicly listed AI cloud peer) and the major cloud platforms.
• Continued senior research-and-engineering talent recruitment.
• Potential expansion into adjacent AI services (training, fine-tuning, agentic AI infrastructure).
• The strategic relationship with NVIDIA and any deeper compute-partnership or commercial alignment.

Sources

• Together AI official site. Company overview and product reference.
• Together AI: RedPajama. Open-source training-data and model release.
• Together AI: StripedHyena 7B. Hybrid architecture release.
• TechCrunch: Together raises $20M to build open source generative AI models. May 2023 seed round.
• TechCrunch: Together lands $102.5M investment. November 2023 Series A.
• SiliconANGLE: Together AI closes on another bumper funding round. 2024 follow-on round and $1.25 billion valuation.
• Together AI on Hugging Face. Open-source model and dataset distribution.