River AI
River AI operates a self-serve, pay-per-token API for LoRA fine-tuning and reinforcement learning on open-source models in the 35B–1T parameter range, serving AI developers and small teams with its own training stack and model family.
- Company typePrivate
- Founded2026
- HeadquartersPalo Alto, United States
- Headcount11–50
- GTM typeB2B
- OfferingSoftware
What River AI does
River AI, Inc. is a developer-facing AI infrastructure company founded in June 2026 by Igor Babuschkin, whose prior experience includes DeepMind, OpenAI, and xAI. The company is headquartered in Palo Alto with an additional office in Austin and operates as a small, elite team whose headcount is not publicly disclosed. No external funding rounds have been disclosed as of the most recent data.
The company's core product is the River API, which exposes LoRA-based fine-tuning and reinforcement learning on open-source language models in the 35 billion to 1 trillion parameter range. River ships its own model family (including the GLM 5.1 and GLM 5.2 releases) and runs a distinctive training pipeline that extends LoRA across all matrices — including Mixture-of-Experts architectures, where it claims "dropless" training capability. The stack exposes named proprietary methods, specifically GRPO and CISPO, alongside a recipe branded "ScaleRL," and is paired with hosted inference on River-managed hardware. Users retain ownership of trained checkpoints, which can also be served on River infrastructure.
River sells exclusively through a self-serve, pay-per-token usage model, initially targeting individual AI developers and builders. The July 2026 introduction of a Teams tier with a shared wallet marks a first step toward group-level accounts. The company's stated long-term vision is to make frontier-quality model customization broadly accessible — ultimately serving individuals seeking a personal AI — but the near-term customer base is the developer and small-team segment. No customer logos, revenue figures, or paying-customer counts have been disclosed, and the product was only released as a v0.1 preview in June 2026.
River AI firmographics
Firmographics- Name
- River AI
- Legal name
- River AI, Inc.
- Website
- https://river.ai
- Company type
- Private
- Founded year
- 2026
- Operating status
- Operating
- Headcount range
- 11–50 employees
- Short description
- River AI operates a self-serve, pay-per-token API for LoRA fine-tuning and reinforcement learning on open-source models in the 35B–1T parameter range, serving AI developers and small teams with its own training stack and model family.
- Ownership category
- akta.pro rank
Where River AI is headquartered
LocationHeadquarters
- HQ city
- Palo Alto
- HQ country
- United States
- HQ region
- North America
Offices2 records
Markets served
River AI business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Technology or R&D, Infrastructure, Personnel, Operations, Marketing or Sales
Revenue model
- Token-based inference and training: Billing is metered on tokens for both inference (prompt and completion) and training. Customers pay per token consumed with no GPU-hour complexity. Checkpoint storage is billed separately at $0.10/GB/month. Longer context lengths are priced on request. This is the sole revenue stream described.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Usage-based | Pay-as-you-go | Qwen3.6-35B MoE — smallest model tier |
| Usage-based | Pay-as-you-go | Qwen3.5-397B MoE — largest model tier |
| Usage-based | Pay-as-you-go | Kimi K2.6 32k — mid-range tier, shorter context |
| Usage-based | Pay-as-you-go | Kimi K2.6 262k — mid-range tier, extended context |
| Usage-based | Pay-as-you-go | GLM 5.2 32k — mid-range tier, shorter context |
| Usage-based | Pay-as-you-go | GLM 5.2 262k — mid-range tier, extended context |
Go-to-market motion1 record
Distribution channels1 record
Marketing channels4 records
River AI product offering
Product offeringCore offering
River AI provides the River API, a Python-based platform that performs LoRA-based fine-tuning and reinforcement learning on open-source foundation models ranging from 35B to 1T parameters (including Qwen, Kimi, and GLM families with MoE architectures). Customers bring their own data, shape the model weights directly, and serve the resulting checkpoints via OpenAI-compatible endpoints under pay-per-token billing, with checkpoint storage billed separately. The River Console web portal provides account, model, and team management on top of the API.
Product overview
River AI is a personal AI company building technology for individual ownership and control of AI intelligence. The core product is the River API (v0.1 preview), which provides LoRA-based fine-tuning and reinforcement learning capabilities on open-source models (Qwen, Kimi, GLM families) ranging from 35B to 397B+ parameters. The platform enables users to train personalized models, have them learn from custom reward signals, and serve them via OpenAI-compatible endpoints. Access is managed through the River Console platform which supports team collaboration with shared-wallet credit management. River's broader mission encompasses building complete personal AI stacks including interfaces, algorithms, infrastructure, and personal hardware.
Differentiator
Problem solved
Functional benefit
Products and services
- River API Python-based API providing LoRA fine-tuning and reinforcement learning on open-source foundation models ranging from 35B to 1T parameters (Qwen, Kimi, GLM families including MoE), with OpenAI-compatible serving endpoints and pay-per-token billing for both inference and training. Built for developers and AI builders who want to train, customize, and deploy task-specific models on their own data.
- River Console Web-based management portal for River AI accounts, providing API key issuance, model and training management, and team collaboration features including shared-wallet credit management. Serves as the self-serve front door for the River API.
Quantifiable outcome
- Under $1,000 for a full production-scale RL training run (~500M completion tokens + ~250M training tokens)
- +2 more outcomes
Companies that use River AI
Customer profileSegments2 records
Ideal customer profiles2 records
River AI technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
AI capability4 records
Feature6 records
River AI partnerships and signals
Strategic signalScale indicators4 records
Recent moves5 records
Expansion highlights5 records
River AI competitors and assessment
Company assessmentDirect peers
- Fireworks AI: Fireworks AI provides fine-tuning and inference APIs for open-source LLMs with pay-per-token pricing and OpenAI-compatible endpoints. Closely overlaps River's product offering and target developer persona.
- RunPod: RunPod offers GPU cloud instances and serverless endpoints for training and serving open-source AI models. Comparable self-serve, pay-as-you-go alternative for model fine-tuning.
- Together AI: Together AI offers a cloud platform for fine-tuning, serving, and inference of open-source models with usage-based pricing. Direct competitor in the open-model fine-tuning API category that River targets.
- Anyscale: Anyscale runs Ray-based AI compute infrastructure for training, fine-tuning, and serving models at scale. Competes for developers who want programmatic, scalable fine-tuning of open models.
- Replicate: Replicate offers a cloud API to run and fine-tune open-source ML models on pay-per-use infrastructure. Direct competitor for developers seeking API access to fine-tuning and inference of open weights.
- Lambda Labs: Lambda provides GPU cloud infrastructure and 1-click fine-tuning clusters for open-source LLMs. Targets the same fine-tuning-on-open-models developer use case with self-serve GPU clusters.
- Modal Labs: Modal provides serverless infrastructure for running and fine-tuning AI models with a Python-native developer experience. Comparable in API-first/self-serve positioning and developer persona.
Broad incumbents
- Hugging Face: Hugging Face is the dominant open-source AI hub offering model hosting, fine-tuning (AutoTrain, Spaces), and inference endpoints. Operates broadly across the same open-model ecosystem River serves.
- OpenAI: OpenAI offers fine-tuning APIs on its proprietary models plus an enterprise-scale inference platform. A broad incumbent whose hosted fine-tuning competes for the same developer/API budget that River targets.
- CoreWeave: CoreWeave is a specialized GPU cloud provider offering large-scale AI training and inference infrastructure. Competes indirectly as the underlying compute layer that hosts fine-tuning workloads similar to River's.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat4 records
Key risks6 records
Key highlights7 records
Customer concentration
River AI social profiles
Digital presenceRiver AI financial estimates
Financial estimateRevenue estimate
Valuation estimate
River AI leadership team
Management profileNumber of profiles
Profiles1 record
River AI funding detail
Funding detailFunding overview
Funding rounds2 records
Investors6 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
River AI M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about River AI
What does River AI do?
River AI provides the River API, a Python-based platform that performs LoRA-based fine-tuning and reinforcement learning on open-source foundation models ranging from 35B to 1T parameters (including Qwen, Kimi, and GLM families with MoE architectures). Customers bring their own data, shape the model weights directly, and serve the resulting checkpoints via OpenAI-compatible endpoints under pay-per-token billing, with checkpoint storage billed separately. The River Console web portal provides account, model, and team management on top of the API.
Is River AI a public or private company?
River AI is a private company. It is classified as venture growth investor backed and is currently operating.
When was River AI founded?
River AI was founded in 2026. It employs 11 to 50 people.
Where is River AI based?
River AI is headquartered in Palo Alto, United States, in the North America region.
How does River AI make money?
One revenue line is on record: token-based inference and training.
Who are River AI's main competitors?
Direct peers on record are Fireworks AI, RunPod, Together AI, Anyscale, Replicate, Lambda Labs and Modal Labs. Broad incumbents are Hugging Face, OpenAI and CoreWeave.
Does River AI have an API?
Yes. LoRA-based fine-tuning and reinforcement learning API enabling training on open-source models from 35B to 1T parameters. Features include: LoRA fine-tuning with low-rank adapters across all weight matrices including MoE experts; RL training with fast weight transfer, sampling-training consistency, and elastic compute; OpenAI-compatible serving endpoint on any checkpoint. Pay-per-token billing model covering inference (prompt and completion) and training costs. Python client library provided. Developer documentation is at river.ai/api.