ZENODO FORM COPY/PASTE — AGASTYA AGI v01 — R1
Prepared: 2026-09-30

IMPORTANT
Do NOT publish until the reserved DOI is inserted into the final files.
Do NOT use "India's first from-scratch foundation model" as an unqualified factual claim in the DOI metadata.
The India-first distinction is a verification target. Sarvam has earlier public from-scratch foundation-model claims.
Use the strong, factual India-developed positioning below.

1) DOI
Select: No, I need one
Then click: Get a DOI now!
Copy the reserved DOI and provide it to the publication owner before uploading the final files.

2) VISIBILITY
Public

3) RESOURCE TYPE
Dataset

4) MAIN TITLE
Agastya AGI v01 Native Brain Evidence Release R1

5) SUBTITLE / ADDITIONAL TITLE
Cryptographically Anchored Provenance for a From-Scratch Sovereign Native Foundation Model Developed in India

6) PUBLICATION DATE
2026-09-30

7) CREATOR
Type: Organization
Name: Rabbit Industries Pvt Ltd
Do not invent a ROR/ORCID. Add an identifier only if Rabbit Industries has a verified one.

8) PUBLISHER
Rabbit Industries Pvt Ltd
Reason: this evidence release was already first made publicly available by Rabbit Industries on 2026-09-30.

9) VERSION
R1
Recommended semantic release label: v1.0.0-evidence-r1

10) LANGUAGE
English

11) DESCRIPTION / ABSTRACT
Agastya AGI v01 is a sovereign native foundation-model program developed in India by Rabbit Industries Pvt Ltd under the Bramha Medha Native INDIA family.

This Evidence Release R1 makes the model's from-scratch lineage externally citable through stable public documentation, cryptographic SHA-256 identifiers, machine-readable provenance manifests, tokenizer metadata, step-0 checkpoint provenance, checkpoint validation, Clean-P0 acceptance evidence, a technical report, a model card, and an independent-verification protocol.

The canonical lineage under evaluation is:

Agastya corpus -> Agastya tokenizer -> Agastya architecture -> RANDOM INITIALIZATION -> Agastya pretraining -> Agastya post-training -> Agastya Native Brain

The published company evidence records a fresh random-initialization step-0 checkpoint, teacher weight ancestry ZERO, third-party weight ancestry false, third-party tokenizer artifact import false, exact unique parameter count of 1,264,715,776, tokenizer vocabulary size of 98,320, and Clean-P0 R2 acceptance PASS.

This release does not redistribute proprietary raw model weights, the raw training corpus, the full tokenizer vocabulary/merges artifact, or private architecture/training source code. Those artifacts are retained for controlled independent verification and are identified through cryptographic hashes where applicable.

Independent third-party verification of the complete lineage remains PENDING. Same-suite frontier benchmark qualification also remains PENDING. No universal or independently verified global performance rank is claimed in R1.

Rabbit Industries' goal is to establish Agastya as a globally competitive, independently verifiable native foundation model from India through reproducible provenance, independent replay, DOI-backed preservation, public model documentation, and frozen same-suite benchmark evaluation.

Canonical evidence:
https://zippri.com/agastya-native-brain-evidence/

Public benchmark profile:
https://zippri.com/agastya-benchmark-profile.html

12) KEYWORDS
Agastya AGI
Bramha Medha
Rabbit Industries
India
Indian artificial intelligence
sovereign AI
native foundation model
from-scratch foundation model
random initialization
model provenance
weight lineage
native tokenizer
checkpoint provenance
cryptographic provenance
AI reproducibility
foundation model
machine learning
independent verification

13) LICENSE
Creative Commons Attribution 4.0 International (CC BY 4.0)
SCOPE: public evidence documentation only.
NOT INCLUDED: model weights, raw corpus, full tokenizer artifact, proprietary architecture/training source code.

14) RELATED IDENTIFIERS
URL: https://zippri.com/agastya-native-brain-evidence/
Relation: Is supplemented by / Is supplement to (choose the closest available relationship)

URL: https://zippri.com/agastya-benchmark-profile.html
Relation: References

URL: https://zippri.com/agastya-competition.html
Relation: References

15) COMMUNITIES
Optional. Leave blank for the first save if no clearly relevant curated AI/ML community is selected.
A community is not required to receive a DOI.

16) FUNDING
Leave blank unless Rabbit Industries has a real grant/funding award that applies to this exact work.

17) INDIA-FIRST POSITIONING
Safe current wording:
"Developed in India by Rabbit Industries Pvt Ltd as a from-scratch sovereign native foundation-model program with publicly released cryptographic provenance evidence."

Verification-target wording:
"Candidate for an independently verified India-first distinction in the narrower category of a cryptographically provenanced native foundation model with zero third-party weight ancestry."

Do NOT state the verification-target wording as an established fact until the prior-art review and independent evaluator confirm it.

18) GLOBAL-BEST POSITIONING
Safe wording:
"Designed and being evaluated for global frontier competitiveness."

Do NOT state "world's best" unless a recognized, directly comparable benchmark or external evaluation establishes the relevant result.
