RFM-1 is credited to Covariant on the DEPLOY registry. Each developer attribution is verified via primary sources.

Is RFM-1 open source?

No. RFM-1 is recorded as proprietary on the DEPLOY registry. Model weights and source are not publicly available.

What is RFM-1's maturity stage?

RFM-1 is at the research stage on the DEPLOY maturity ladder. Research stage means active development without commercial deployments on file.

Which robots run on RFM-1?

No robot models on the DEPLOY registry are recorded as running RFM-1. DEPLOY wires brain-to-model connections only when the wiring is verifiable from primary sources; absence may reflect pre-deployment or unverified manufacturer claims.

Brain

RFM-1

Covariant (founded 2017 in the Berkeley area by Pieter Abbeel, Peter Chen, Rocky Duan, and Tianhao Zhang) built RFM-1, the Robotics Foundation Model announced in March 2024: a roughly 8-billion-parameter multimodal any-to-any sequence model trained on text, images, video, robot actions, and sensor readings, used to power warehouse pick-and-place robots. The registry records the brain at commercial maturity to reflect genuine deployed wiring: pre-acquisition, the Covariant Brain ran on real robotic arms at named customers including Otto Group, Radial, and McKesson, not just in demonstrations. Its corporate state is a key big-tech-absorption datapoint: in a reverse-acquihire that closed on August 30, 2024, Amazon hired founders Abbeel, Chen, and Duan along with about a quarter of staff and took a non-exclusive license to Covariant's models, reportedly for $380 million plus a $20 million payment per an FTC whistleblower complaint, with Covariant continuing as a diminished but independent company (NOT an Amazon subsidiary) under CEO Ted Stinson and remaining co-founder Tianhao Zhang, retaining its IP, while Amazon is a non-exclusive licensee rather than the brain developer; the Covariant Brain remains deployed via the KNAPP Pick-it-Easy OEM channel at customers such as McKesson, Wurth, GXO, and Brodrene Dahl. RFM-1 is proprietary with no peer-reviewed paper, model card, or released checkpoints, the human-like-reasoning framing is marketing, and current independent deployments and whether RFM-1 itself runs inside Amazon are not verified.

Foundation model · Maturity: Research · Closed

Machine-readable surfaces

Markdown mirror: /brains/rfm-1.md
RSS feed: /brains/rfm-1/feed.xml
JSON-LD: embedded in this page’s head
REST API: /v1/brains/c3682994-51ee-4694-9112-1d14b46271af
Revision history: /brains/rfm-1/history
Data documentation: /data
Query this programmatically: Deploy MCP

Architecture

Multimodal any-to-any autoregressive sequence model (~8B params); next-token prediction over text/images/video/robot-actions/sensors.

Key facts

Model: RFM-1 (frozen): Robotics Foundation Model, announced Mar 2024 at MODEX: ~8B-param multimodal any-to-any sequence model (text/images/video/robot-actions/sensors). Proprietary. NO successor (no 'RFM-2') and NO arXiv/peer-reviewed paper ever - disclosed only via blog/press. The absence of any successor model since the Aug 2024 team transition is itself the finding.
Corporate state (precise; load-bearing): Covariant is a CONTINUING-BUT-DIMINISHED independent company, NOT an Amazon subsidiary. Aug 30 2024 reverse-acquihire: Amazon hired 3 of 4 co-founders (Peter Chen, Pieter Abbeel, Rocky Duan) + ~25% of staff and took a NON-EXCLUSIVE license to Covariant's models (Covariant RETAINED the IP; nothing was sold). The 4th co-founder, Tianhao Zhang, REMAINED; former COO Ted Stinson became CEO. Pieter Abbeel now leads frontier-model research within Amazon's AGI org (Dec 2025; secondary-sourced).
Deployment evidence (verified-vs-claimed): Covariant-proper (VERIFIED, via OEM channel): the Covariant Brain remains deployed through the KNAPP 'Pick-it-Easy Robot' relationship at ~26 KNAPP customers (incl. McKesson, Wurth, GXO, Brodrene Dahl); ABB/Radial/Otto Group are historical. No verified net-new customer wins after Aug 2024. Amazon side (CLAIMED, not verified): Amazon credits Covariant founders/models/infrastructure with accelerating its 'Blue Jay' multi-arm system (unveiled Oct 2025) - but no source confirms RFM-1 ITSELF runs in Amazon production, and Blue Jay was DISCONTINUED Feb 2026 after missing targets.
Cap-flag (allegation-sourced + a false claim to avoid): The $380M + $20M deal value and the 'zombie company existing only on paper' characterization come from a 2025 FTC/SEC/DOJ WHISTLEBLOWER COMPLAINT, not confirmed by either company - treat as allegation; Covariant's own framing (continuing to serve customers) contradicts it, surface both. No peer-reviewed paper, model card, or released checkpoints; 'human-like reasoning' is marketing. FALSE claim to avoid: one aggregator says 'Covariant was acquired by Intrinsic, an Alphabet subsidiary' - this is FALSE (conflation; it was the Amazon reverse-acquihire). Do not ingest.
Maturity = research (2026-06-04; corporate-state vs model-state distinction): maturityStage changed commercial->research. Rationale (verified-vs-claimed applied recursively at corporate-vs-model granularity): the MODEL RFM-1 specifically has NO verified production deployment - Amazon's use of RFM-1 is claimed-not-verified, and the Blue Jay system built with Covariant tech was discontinued Feb 2026. The COMPANY Covariant does have verified commercial deployment, but that is via the legacy Covariant Brain through the KNAPP Pick-it-Easy OEM channel, NOT RFM-1. Both are true simultaneously: Covariant-as-company = commercial (legacy Brain via KNAPP); RFM-1-as-model = research (no verified production). The brain maturityStage tracks the MODEL, so it is research. Corporate state unchanged: continuing-but-diminished independent company under CEO Ted Stinson (NOT an Amazon subsidiary; non-exclusive license; IP retained).

Developed by (1)

Covariant

Common questions

What is RFM-1?: Covariant (founded 2017 in the Berkeley area by Pieter Abbeel, Peter Chen, Rocky Duan, and Tianhao Zhang) built RFM-1, the Robotics Foundation Model announced in March 2024: a roughly 8-billion-parameter multimodal any-to-any sequence model trained on text, images, video, robot actions, and sensor readings, used to power warehouse pick-and-place robots. The registry records the brain at commercial maturity to reflect genuine deployed wiring: pre-acquisition, the Covariant Brain ran on real robotic arms at named customers including Otto Group, Radial, and McKesson, not just in demonstrations. Its corporate state is a key big-tech-absorption datapoint: in a reverse-acquihire that closed on August 30, 2024, Amazon hired founders Abbeel, Chen, and Duan along with about a quarter of staff and took a non-exclusive license to Covariant's models, reportedly for $380 million plus a $20 million payment per an FTC whistleblower complaint, with Covariant continuing as a diminished but independent company (NOT an Amazon subsidiary) under CEO Ted Stinson and remaining co-founder Tianhao Zhang, retaining its IP, while Amazon is a non-exclusive licensee rather than the brain developer; the Covariant Brain remains deployed via the KNAPP Pick-it-Easy OEM channel at customers such as McKesson, Wurth, GXO, and Brodrene Dahl. RFM-1 is proprietary with no peer-reviewed paper, model card, or released checkpoints, the human-like-reasoning framing is marketing, and current independent deployments and whether RFM-1 itself runs inside Amazon are not verified.
Who developed RFM-1?: RFM-1 is credited to Covariant on the DEPLOY registry. Each developer attribution is verified via primary sources.
Is RFM-1 open source?: No. RFM-1 is recorded as proprietary on the DEPLOY registry. Model weights and source are not publicly available.
What type of AI is RFM-1?: RFM-1 is a foundation model, built on a Multimodal any-to-any autoregressive sequence model (~8B params); next-token prediction over text/images/video/robot-actions/sensors. architecture on the DEPLOY registry.
What is RFM-1's maturity stage?: RFM-1 is at the research stage on the DEPLOY maturity ladder. Research stage means active development without commercial deployments on file.
Which robots run on RFM-1?: No robot models on the DEPLOY registry are recorded as running RFM-1. DEPLOY wires brain-to-model connections only when the wiring is verifiable from primary sources; absence may reflect pre-deployment or unverified manufacturer claims.

Sources (10)

Methodology: Verified · 10 sources (2 primary) · last reviewed 2026-06-04

Verification posture

Verified

High confidence

Review state

Stable

Last reviewed 2026-06-04

Maturity + lifecycle

Maturity stage: research

Sources by quality tier

2: primary-company-ir; Company IR disclosure
2: unclassified; Unclassified source
2: secondary-established-publication; Established publication
2: secondary-trade-publication; Trade publication
1: secondary-industry-publication; Industry publication
1: knowledge-base; Knowledge base

The framework is documented at /methodology. Corrections at /corrections. Reviewer: DEPLOY editorial team.

Canonical ID c3682994-51ee-4694-9112-1d14b46271af