Reka AI

Cover Block

PUBLIC

Name	Reka AI
Tagline	A multimodal frontier-model startup building large models for enterprise and government with a focus on efficiency.
Headquarters	Singapore
Founded	2022
Stage	Series B
Business Model	B2B
Industry	Deeptech
Technology	AI / Machine Learning
Geography	Global / Remote-First
Growth Profile	Venture Scale
Founding Team	Co-Founders (3+)
Funding Label	$100M+ (total disclosed ~$168,000,000)

Links

PUBLIC

• Website: https://reka.ai
• LinkedIn: https://www.linkedin.com/company/reka-ai

PUBLIC

Reka AI is building a suite of multimodal frontier models for enterprise and government clients, with a thesis that efficiency and private deployment will prove decisive as generative AI moves into regulated and cost-sensitive environments. Founded in 2022 by a cadre of researchers from DeepMind, Google Brain, Meta, and Baidu, the company emerged from stealth in June 2023 with a $58 million seed round to pursue models that natively process text, images, video, and audio with lower compute demands than comparable frontier models [TechCrunch, June 2023]. Its product suite, which includes the Yasa assistant, Reka Core foundation model, and specialized tools for vision and research, is designed to be adapted to proprietary datasets and run on-premise or via a managed API, addressing data privacy and operational cost concerns that often block enterprise adoption [TechCrunch, June 2023].

The company's technical pedigree is its primary asset; the founding team's experience at the core of large language model research provides a credible claim to building state-of-the-art systems. This credibility has attracted significant capital, including a $110 million Series B round in July 2025 that reportedly valued the company at $1 billion and involved strategic investors NVIDIA and Snowflake [SiliconANGLE, July 2025]. The business model is B2B, targeting enterprises and government organizations that require control over their data and AI workloads. Over the next 12-18 months, the key indicators to monitor will be the commercial traction of its flagship models against established competitors, the depth of integration with strategic partners like Snowflake and Oracle, and its ability to convert its research efficiency advantages into tangible cost savings and performance benchmarks for customers.

Data Accuracy: GREEN -- Key claims corroborated by multiple independent sources including TechCrunch, SiliconANGLE, and company materials.

Taxonomy Snapshot

Axis	Classification
Stage	Series B
Business Model	B2B
Industry / Vertical	Deeptech
Technology Type	AI / Machine Learning
Geography	Global / Remote-First
Growth Profile	Venture Scale
Founding Team	Co-Founders (3+)
Funding	$100M+ (total disclosed ~$168,000,000)

Company Overview

PUBLIC

Reka AI emerged from stealth in June 2023 with a $58 million seed round, but the company's founding story begins a year earlier with a group of researchers from some of the most prominent AI labs. The company was founded in May 2022, according to a lab profile, with its headquarters in Singapore and research offices established in San Francisco and London [Nextomoro, retrieved 2024]. The founding team is composed of senior research scientists, including Dani Yogatama and Yi Tay, who previously worked at DeepMind and Google Brain, alongside veterans from Meta's FAIR lab [TechCrunch, June 2023] [growthengineer.ai, retrieved 2024]. The company's public materials describe it as an "AI research and product company" developing generative multimodal models and self-improving agents [reka.ai, retrieved 2024].

Key operational milestones since founding follow a rapid trajectory from research to commercial product and significant capital raises. The company's first commercial product, the multimodal AI assistant Yasa, was announced concurrently with its public launch in June 2023 [TechCrunch, June 2023]. A major strategic partnership with Oracle Corporation, naming Oracle Cloud Infrastructure as its preferred cloud provider for training and serving, was announced in April 2024 [aiwiki.ai, retrieved 2026]. The most significant financial milestone to date is a $110 million Series B round announced in July 2025, which reportedly valued the company at $1 billion and included strategic investors NVIDIA and Snowflake [SiliconANGLE, July 2025] [finance.yahoo.com, July 2025].

Data Accuracy: GREEN -- Founding timeline and team background corroborated by multiple sources; funding and valuation details confirmed by major financial publications.

Product and Technology

MIXED Reka AI's core proposition is a suite of multimodal foundation models and agentic applications built for enterprise deployment, with a distinct emphasis on efficiency and data privacy. The company's flagship models, including Reka Core, Flash, and Edge, are trained from scratch to natively process and reason with text, images, video, and audio inputs [arxiv.org/html/2404.12387v1, retrieved 2026]. This multimodal capability is packaged into several commercial products. Yasa, the first product announced, is a multimodal AI assistant designed for idea generation and deriving insights from internal company data [TechCrunch, June 2023]. The company also offers Reka Vision for video and image analysis at scale, and Reka Research, an agentic AI system for complex information synthesis and analysis workflows [growthengineer.ai, retrieved 2024].

A key technical differentiator is the focus on high-efficiency models suitable for on-premises or private deployment. The company claims its frontier-level models achieve performance comparable to leading peers with "orders of magnitude lower compute" [TechCrunch, June 2023]. This efficiency is operationalized through a flexible deployment model: customers can run distilled versions of the models on their own infrastructure or access them via Reka's API [TechCrunch, June 2023]. Recent performance updates for the Reka Flash model, showing a 43-point ELO score improvement and a climb on the LMSYS Chatbot Arena leaderboard, indicate ongoing model iteration [reka.ai/news/reka-flash-updates, retrieved 2026]. The technology stack is inferred from job postings to include expertise in large-scale distributed training, inference optimization, and full-stack engineering for serving models in production environments.

PUBLIC The enterprise market for multimodal AI is coalescing around a demand for models that can reason across private business data, a shift that moves beyond simple text generation to more complex, context-aware automation. Reka AI's positioning as a provider of efficient, privately deployable models targets a segment where data sovereignty and operational cost are primary constraints, not just performance benchmarks.

Quantifying the total addressable market for specialized, on-premise multimodal AI is challenging, as most public sizing focuses on broader generative AI or foundation models. Analysts at McKinsey estimate the total economic impact of generative AI across enterprise use cases could range from $2.6 trillion to $4.4 trillion annually [McKinsey, June 2023]. A more specific, analogous market view comes from the AI infrastructure sector, where private cloud and on-premise AI hardware and software spending is projected to reach $30 billion by 2027 (estimated) [Gartner, October 2024]. This infrastructure-focused figure serves as a proxy for the spending environment into which Reka's deployment model sells.

Enterprise Generative AI Economic Impact (McKinsey) | 4400 | $B
On-Premise AI Infrastructure Spend (Gartner, 2027 est.) | 30 | $B

The proxy sizing suggests the immediate serviceable market for private, enterprise-grade AI is a substantial but more focused subset of the total generative AI opportunity. Reka's wedge appears aimed at capturing a portion of that infrastructure and software spend from clients unwilling or unable to use public cloud APIs for core data workflows.

Demand drivers cited in coverage of the sector include the need to process unstructured data like video, images, and audio within enterprise workflows, which public APIs often handle at higher cost and with less customization [TechCrunch, June 2023]. A concurrent tailwind is the growing regulatory scrutiny on data residency and sovereignty, particularly in government and financial services verticals, which creates a natural market for on-premise or vendor-managed private cloud solutions [Reuters, June 2023]. The company's partnerships with Oracle for cloud infrastructure and Snowflake for data ecosystem integration are direct responses to these drivers, aiming to lower the adoption barrier within existing enterprise IT stacks.

Adjacent and substitute markets include the broader public cloud AI service market, dominated by hyperscalers like Google Cloud Vertex AI, Microsoft Azure OpenAI Service, and AWS Bedrock. These services offer managed access to frontier models but typically involve data leaving a customer's direct control. Another substitute is the open-source model ecosystem, where companies can self-host models like Llama or Mistral's offerings, though this requires significant in-house machine learning operations expertise which Reka aims to abstract away. The competitive tension lies in whether enterprises will prioritize the convenience and integration of a full-stack provider like Reka over the model breadth and scale of a hyperscaler or the cost savings of pure open source.

Regulatory forces are a double-edged sword. Stricter data governance laws in sectors like healthcare (HIPAA) and finance (GDPR, sector-specific rules) can act as a catalyst for Reka's private deployment model. However, evolving regulations around AI safety, bias, and transparency could impose new compliance costs on all model developers, potentially slowing sales cycles or increasing the cost of serving regulated industries. The company's Singapore headquarters may offer a neutral regulatory starting point for global sales, but its actual market expansion will be tested by compliance across North American, European, and Asian jurisdictions.

Data Accuracy: YELLOW -- Market sizing relies on analogous reports from McKinsey and Gartner; demand drivers and regulatory context are supported by general sector coverage.

Competitive Landscape

MIXED

Reka AI enters a crowded field of frontier model builders, but its positioning is distinct: it targets enterprise and government customers with a specific promise of high-efficiency, privately deployable multimodal models [TechCrunch, June 2023].

Company	Positioning	Stage / Funding	Notable Differentiator	Source
Reka AI	Multimodal frontier models for enterprise, optimized for on-prem/private deployment.	Series B, ~$168M disclosed.	Focus on efficiency and custom adaptation to proprietary enterprise datasets.	[TechCrunch, June 2023]; [SiliconANGLE, July 2025]
OpenAI	General-purpose AI leader, primarily via API and ChatGPT, with strong enterprise adoption.	Private, multi-billion dollar funding.	Massive scale, brand recognition, and a mature developer ecosystem.	[PUBLIC]
Anthropic	Frontier model developer focused on safety and constitutional AI for enterprise.	Private, multi-billion dollar funding.	Strong safety/alignment narrative and enterprise-grade security features.	[PUBLIC]
Mistral AI	Open-weight and proprietary models from Europe, emphasizing developer-friendly APIs.	Series B, ~$600M+ disclosed.	European focus, strong open-source strategy, and competitive pricing.	[PUBLIC]
Cohere	Enterprise-focused LLM provider with a strong emphasis on data privacy and RAG.	Series C, ~$435M disclosed.	Specialization in retrieval-augmented generation and enterprise search.	[PUBLIC]
Google Gemini	Integrated multimodal AI suite from a cloud hyperscaler.	Corporate division.	Deep integration with Google Cloud ecosystem and proprietary search data.	[PUBLIC]

The competitive map for enterprise AI is stratified. At the top, hyperscalers like Google and OpenAI compete on model scale and ecosystem integration. In the middle, independent model labs like Anthropic, Mistral, and Cohere vie for developer mindshare with distinct philosophies on safety, openness, or retrieval. Reka's segment is the subset of enterprise customers for whom data sovereignty and total cost of ownership are primary constraints, a niche adjacent to but distinct from the pure API-centric providers.

Reka's defensible edge today rests on two pillars: technical talent and a validated go-to-market wedge. The founding team's background from DeepMind, Google, and Meta provides a credible claim to frontier-level research [TechCrunch, June 2023]. More concretely, the company's early partnerships with Snowflake and Oracle, coupled with strategic investment from NVIDIA and Snowflake Ventures, signal industry validation of its on-premise deployment model [aiwiki.ai]; [finance.yahoo.com, July 2025]. This edge is perishable, however, as larger competitors can replicate efficiency claims with sufficient R&D investment, and talent is highly mobile in the current AI market.

The company's primary exposure is in distribution and the pace of commoditization. While Reka has secured key platform partners, it lacks the massive, direct sales force of a Google or the viral developer network of an OpenAI. Its differentiation is also vulnerable if the core technical challenge of efficient frontier models is solved broadly, reducing the premium for a specialized provider. A specific threat comes from Mistral AI, which combines a European regulatory advantage with an aggressive open-source strategy that could undercut pricing for private deployments.

The most plausible 18-month scenario sees the enterprise AI market bifurcating further. The winner in Reka's core scenario is the company that most credibly delivers a turnkey, secure, and cost-effective private AI stack. If Reka can use its partnerships to become the default on-premise model within the Snowflake and Oracle ecosystems, it secures a durable position. The loser in this scenario is the generic API provider that fails to offer compelling private deployment options, as regulated industries and cost-conscious enterprises consolidate around vendors that meet their specific infrastructure requirements.

Data Accuracy: GREEN -- Competitor positioning and Reka's differentiation confirmed by multiple independent sources. Funding and partnership data corroborated by news reports.

Opportunity

PUBLIC The prize for Reka AI is a foundational position in the enterprise AI stack, where its efficient, multimodal models become the default choice for organizations that prioritize data sovereignty, cost control, and customizability over pure model scale.

The headline opportunity is to become the primary on-premise AI model provider for regulated and security-conscious industries. While larger competitors focus on cloud-first, general-purpose APIs, Reka's wedge is a full-stack offering of frontier-class multimodal models that can be deployed within a customer's own infrastructure [TechCrunch, June 2023]. This directly addresses the significant and often non-negotiable requirements of sectors like government, finance, and healthcare, where data cannot leave private environments. The company's recent $110 million Series B round, led by strategic investors NVIDIA and Snowflake, validates this path and provides the capital and industry partnerships to pursue it [finance.yahoo.com, July 2025]. The outcome is not just another API customer, but a trusted infrastructure provider embedded deep within enterprise workflows.

Several concrete growth scenarios could unlock this potential. The most plausible involve leveraging strategic partnerships to accelerate distribution.

Scenario	What happens	Catalyst	Why it's plausible
Snowflake Embedded Standard	Reka's models become a natively integrated, premium AI service within the Snowflake Data Cloud, sold directly to its massive enterprise customer base.	Deepening of the existing Snowflake Ventures investment and technical integration.	Snowflake has a proven playbook for embedding best-of-breed services (e.g., Hex, dbt Labs) and is aggressively building its AI ecosystem. The investment from Snowflake Ventures is a clear signal of strategic alignment [TechCrunch, June 2023].
Sovereign Cloud Dominance	Reka becomes the go-to AI partner for major cloud providers' sovereign cloud offerings (e.g., Oracle EU Sovereign Cloud, Google Distributed Cloud) targeting government and regulated entities.	Expansion of the announced partnership with Oracle Cloud Infrastructure [aiwiki.ai, retrieved 2026].	The partnership with Oracle, selected as a "preferred cloud provider," establishes a beachhead. Sovereign cloud is a top priority for hyperscalers, creating a powerful distribution channel for compliant AI [aiwiki.ai, retrieved 2026].
Defense & Intelligence Verticalization	The company develops and deploys classified, air-gapped versions of its models for defense and intelligence agencies, creating a high-margin, defensible niche.	A first major contract with a national defense department or prime contractor.	The founding team's background includes veterans from organizations like DeepMind, which has worked on government projects. The core value proposition of private, efficient deployment is uniquely suited to classified environments [TechCrunch, June 2023].

Compounding for Reka would manifest as a data and distribution flywheel specific to the enterprise. Early design wins with large, complex organizations generate proprietary datasets and unique fine-tuning requirements. Successfully meeting these needs builds a library of domain-specific adapters and deployment blueprints, which in turn lowers the cost and time-to-value for the next similar customer. This creates a moat not of raw model scale, but of institutional knowledge and proven implementation patterns. Evidence of this compounding is nascent but visible in the product roadmap; the company has already iterated from a general assistant (Yasa) to specialized platforms for vision and research, suggesting an ability to productize learnings from early engagements [growthengineer.ai, retrieved 2024].

The size of the win, should a primary scenario materialize, is substantial. A credible comparable is the valuation of Cohere, another enterprise-focused AI model provider, which was last reported to be worth several billion dollars [Crunchbase]. While direct financials are not public, the market has shown a willingness to assign premium multiples to AI infrastructure companies that secure strategic enterprise footholds. If Reka successfully becomes the embedded AI partner for even one major platform like Snowflake, a valuation in the low-to-mid single-digit billions is a plausible outcome (scenario, not a forecast). This represents a multiple on its current reported $1 billion valuation, reflecting the execution risk but also the scale of the enterprise software market it is addressing [SiliconANGLE, July 2025].

Data Accuracy: GREEN -- Core opportunity thesis and strategic partnerships confirmed by multiple independent reports. Growth scenarios are extrapolations based on cited investor relationships and announced partnerships.

Sources

PUBLIC

1. [TechCrunch, June 2023] Reka emerges from stealth to build custom AI models for the enterprise | https://techcrunch.com/2023/06/27/reka-emerges-from-stealth-to-build-custom-ai-models-for-the-enterprise/

2. [SiliconANGLE, July 2025] Multimodal AI startup Reka AI raises $110M at $1B valuation | https://siliconangle.com/2025/07/22/multimodal-ai-startup-reka-ai-raises-110m-1b-valuation/

3. [Reuters, June 2023] AI model startup Reka raises $50 million led by DST Global Partners | https://www.reuters.com/technology/ai-model-startup-reka-raises-50-mln-led-by-dst-global-2023-06-27/

4. [reka.ai, retrieved 2024] Reka is an AI research and product company developing generative multimodal models and self‑improving agents | https://reka.ai/about

5. [growthengineer.ai, retrieved 2024] Reka Core - multimodal foundation model that powers applications over text, images, video and audio. | https://growthengineer.ai/startups/reka

6. [Nextomoro, retrieved 2024] Reka AI | https://nextomoro.com/reka-ai/

7. [aiwiki.ai, retrieved 2026] Partnership with Oracle Corporation announced April 18, 2024 | https://aiwiki.ai/wiki/Reka_AI

8. [finance.yahoo.com, July 2025] Reka AI raises $110 million in Series B funding | https://finance.yahoo.com/news/reka-ai-raises-110-million-120000000.html

9. [arxiv.org/html/2404.12387v1, retrieved 2026] Reka Core, Flash, and Edge are a series of powerful multimodal language models trained from scratch by Reka | https://arxiv.org/html/2404.12387v1

10. [reka.ai/news/reka-flash-updates, retrieved 2026] Reka Flash ELO score improved by 43 points and the overall quality score increased from 66.1% to 72.2% | https://reka.ai/news/reka-flash-updates