Organization
Reka
Model developer · Research lab · Inference provider · San Francisco, California
Reka is an AI research lab that develops multimodal models (the Spark, Edge, Flash, and Core lines) that process video, images, audio, and text, and offers them through an API. It also runs Reka Cloud, a platform for serverless inference, dedicated GPU capacity, and model fine-tuning, and it has released some models, including Reka Flash 3 and 3.1, as open weights.3456
Using this organization’s offerings
The detailed audience, self-hosting, pricing, and restrictions review has not yet been completed. Consult the sourced product records below and the publisher’s current terms. Missing detail means unknown, not unavailable.
Key facts
- Ownership (separate from openness)
- Unknown
Recorded classification; fact-level ownership verification pending.
- Legal form
- Unknown
- Founded
- Unknown
- Sectors
- Open models, Cloud, Research
What is and isn’t public
Reka offers most of its models through its hosted API. It has published some models as open weights on Hugging Face, including Reka Flash 3 and 3.1 under the Apache License 2.0; Reka Edge 2603 is published under a separate Reka license file rather than a standard open license. Openness is assessed per release in this catalog.4101211
Products
Hosted model products
Labeled Hosted product/API — this describes how a documented product is delivered (hosted access). It does not determine whether underlying weights or code are open; assess the specific release and its licenses.
| Product | Kind | Documented delivery and access | Reviewed |
|---|---|---|---|
| Reka API (external site: docs.reka.ai) | Hosted model API | Hosted API for Reka's own models, authenticated with an API key, covering multimodal chat, video understanding, and generation of short clips from long videos.43 | Oct 2, 2026 |
Other products
| Product | Kind | Documented delivery and access | Reviewed |
|---|---|---|---|
| Reka Cloud (external site: reka.ai) | Cloud platform | Platform with an OpenAI-compatible endpoint for serverless inference on open models from several developers, dedicated NVIDIA GPU capacity, and fine-tuning and reinforcement learning services; self-serve signup at developer.reka.ai or through Reka's sales team.5 | Oct 2, 2026 |
Cataloged artifacts
Model families
- Reka FlashModel family
Model releases
- Reka Flash 3.1Model releaseAvailability: PublicModel-disclosure tier (USASI rubric v0.2): Open-weight
Notable documented facts
- In June 2026 Reka announced that it had joined forces with Moonvalley, adding a team of Moonvalley researchers and engineers to Reka; the announcement calls this a merger and describes Reka as headquartered in San Francisco.6
- Reka released Reka Flash 3, a 21B-parameter model it says was pretrained from scratch, as open weights under the Apache License 2.0 in March 2025, followed by Reka Flash 3.1 in July 2025 (see the Reka Flash records).89
Open positions
This job source has not been successfully verified recently. Check the employer’s careers page for current openings.
What this catalog does not know
- Founding date: not documented in the sources reviewed.
Have a primary source? How to report a correction.
U.S. eligibility
Eligible · basis: U.S. headquarters
Reka's privacy policy (effective June 2026) covers Reka AI, Inc. and its subsidiaries, states "We are headquartered in the United States," and gives a Sunnyvale, California mailing address; its terms of use (updated September 2026) name Reka AI, Inc. at the same address and choose California law. Reka's June 2026 announcement describes it as headquartered in San Francisco. The privacy policy's list of data recipients mentions a "corporate parent" without naming one; no parent company was identified in the pages reviewed, and Reka's own pages present Reka AI, Inc. as the operating company. Reka's open roles are listed in the U.S., the U.K., Singapore, and remote; that does not change the documented U.S. headquarters.2167
Sources
This listing is not an endorsement or a federal approval.