CatalogWantedMirrorsLedgerOperating StatementDocsStudio statusCard system
alt.freeweights.nemotron / sft-repro-thinking-step630-nemotron-terminal-step1888
From: laion <huggingface.co/laion>Newsgroups: alt.freeweights.nemotronSubject:

sft-repro-thinking-step630-nemotron-terminal-step1888

Date: 2026-11-15Commit: 0ef7531Signed: 2DC28CC7B26A283B ✓

Text generation model with 67.1B parameters. Safetensors weights, 134 GB in all. Published by laion under Apache-2.0.

67.1B params134GBsft-repro-thinking-step630-nemotron-terminal-step1888safetensorspermissive
FactsParams67.1BSize134 GBFormatssafetensorsTasktext generationLicenseApache-2.0From sourcepermissive
$ fw get laion/sft-repro-thinking-step630-nemotron-terminal-step1888
# manifest sha256 28dee073d4a4… · 134 GB · 47 files
$ fw verify ./sft-repro-thinking-step630-nemotron-terminal-step1888 laion/sft-repro-thinking-step630-nemotron-terminal-step1888
From sourceHosted (signed)laion/sft-repro-thinking-step630-nemotron-terminal-step1888Last check: 2026-11-15 04:20 UTC
Get help

Get help: custom fine-tune or private deploy. We review your setup before we quote.

Works with

Run24Fine-tune29Convert15Evaluate13Scan13Apps58
OverviewModel cardFilesUse itLicenseProvenanceModel tree

Overview

Size
134 GB
Weights
134 GB (safetensors)
Parameters
67.1B
Formats
safetensors
Task
text generation
License
Apache-2.0 (permissive)
Files
47, each with a sha256
Publisher
laion
Added
2026-11-15
Section
alt.freeweights.nemotron
Lineage

Original release from the publisher.

Where these bytes come from
Publisher at the pinned commit→You

Sources are tried in order. A file counts only when its size and sha256 match the manifest. Sources are checked when you ask for them, not on a timer.

Sources
BadgeHostLast checkResult
From sourcehuggingface.co2026-11-15 04:20 UTCpassed

Model card

From the publisher at commit 0ef7531. Shown as plain text with links; images open at the publisher.

Grug 67B Nemotron Terminal SFT reproduction#

This is the BF16 Hugging Face/vLLM export evaluated by the Marin experiment sft-repro-nemotron-terminal-on-thinking-ckpt.

Training#

  • Initialization: Marin Grug 67B stage-2 thinking checkpoint, step 630
  • Data: nvidia/Nemotron-Terminal-Corpus at revision a1667c4, using Marin's documents/nemotron_terminal_full@2026.06.28 selection and tokenized/nemotron_terminal_full-chat-7adc64@2026.07.17
  • Steps: 1,888
  • Sequence length: 32,768
  • Global batch size: 64
  • Peak learning rate: 5e-6, AdamH/Adam with cosine decay, 3% warmup, and 0.1 minimum learning-rate ratio
  • Training geometry: 64 H100 GPUs; replica=1, data=8, expert=8, model=1
  • Marin source: e502edd70c

The training run completed with a final reported loss of 0.501. The export was generated from the final step-1,888 checkpoint and preserves the model's native 32,768-token context limit.

Evaluation#

The model was evaluated on the 300-trial OpenThoughts-TBLite reproduction campaign with the Terminus-2 harness. The experimental controls use three attempts per task, temperature 0.6, and an agent context budget of 24,576 input plus 8,192 output tokens. Evaluation results are tracked separately from this weight artifact.

Provenance#

The immutable source export is:

s3://marin-us-east-02a/marin/grug/sft_repro_thinking_step630_nemotron_terminal_steps1888/2026.08.11.1/exports/step-1888/hf-bf16-vllm/

Files

47 files · 134 GBWeights 134 GB (safetensors). Sizes and sha256 from the signed manifest.
Show all 47 files with their sha256
FileSizesha256
.gitattributes1.57 KB34448b82c17d60fec9b65b1f093c115ddbaadc04beb1b0140b6bfed2e012a930
README.md1.57 KB0271d5a32ba074958b15cfd425ae571918344b76599a64d477c7650111b46bc1
chat_template.jinja5.11 KB414d8559d363292c811e624fe2db04d52647ca5bd89acc53e98bd4c4f371302a
config.json1.06 KBf1d693de07452ccee0855700642ef903efdf1a8e321bdc1436806bf7d393a8d2
generation_config.json82 B838158c033c183c9c7e254c03b289d2f86a3cfe1e0878cb9bbf3b6a0c1a18ea8
model-00001-of-00039.safetensors4.67 GB356d47d6f6b86f4d5ecc07e826b45588d4337ddbb5de975c933e373ec5d2fccf
model-00002-of-00039.safetensors3.43 GBe4f01b1c1f33707fab3a7e163c91cf588252a4284972209f7f2ffbbba8460ddc
model-00003-of-00039.safetensors3.43 GBfe8dc2a60919c12b93669422e07ac48616416ce514b3ad123766745c612e113a
model-00004-of-00039.safetensors3.36 GB2b4d18fc635d4fb312d171fc3d63595a4de9c7b373add7d19a56c77843bf338d
model-00005-of-00039.safetensors3.43 GB9422b47b82a79f67394585fcf342130bc24bcbdc6b6772bcf9b3be7772798b76
model-00006-of-00039.safetensors3.43 GBb38041b421cc87904df20f4cc47919b9d383ed0606308860f3f853fed8a5e11e
model-00007-of-00039.safetensors3.36 GBf2f070e082577f3a3981e06d7306bcd3b198724af01d72b70262044341ff8c92
model-00008-of-00039.safetensors3.43 GBb8a32ac9198838116ba804e72c97e0d59bdd7229ac88c94cbe3b64de0594f67c
model-00009-of-00039.safetensors3.43 GB3253998bc6718be6245f617e1b7b032dcd6e8a585a5357e3de14bdd2b8101040
model-00010-of-00039.safetensors3.36 GBc00f2fd853c08fb7e8bd121642cc52a2696eefe123f1b8766978dd440425f973
model-00011-of-00039.safetensors3.43 GB0f2934f62bbbfb26373c1e62dea00f097b529dfffc56fed53832f26296a4eb3b
model-00012-of-00039.safetensors3.43 GB694b2c5e4c9af77ac92f7528e84a7125982ffb557d303b6defd91d4add28d7df
model-00013-of-00039.safetensors3.36 GBa8575eead79e591de1014c4ffaa13dcdc075891223deee6622e95d03a450afd4
model-00014-of-00039.safetensors3.43 GBfcd249e0870e6d6125fddf43043e89544035ec460ea936ea36b4c25c15968fd7
model-00015-of-00039.safetensors3.43 GBb7b75b98c69f0e3be7cd4374abe26ce45ca65c51cf5fdf97144b8fced42b6356
model-00016-of-00039.safetensors3.36 GBe1919194c64502ca44f0a8d3f682184c37ee6163f4d6814bb4610551b7517d1e
model-00017-of-00039.safetensors3.43 GB34f4725d9f43e88f7326f08152f6c133f1db8f57d14970140e473959a9347901
model-00018-of-00039.safetensors3.43 GB8f68d9cecf53eb9973c0a9d5bc5e72398ee300d87097994e16cff3c4210ea146
model-00019-of-00039.safetensors3.36 GB693bbb4e9c4337c99fb93e3ffa595d8a22de7d2e60b9498f94a04c4ff177919c
model-00020-of-00039.safetensors3.43 GBee377a90e4cb729349e38e88b1370f1c5cf6d9c45fc42bb3c82d2d8ad1e82592
model-00021-of-00039.safetensors3.43 GBbe002a44071dee7ea0433bff433dd13acc8ca31680c7ba048cc4c54ae913abbd
model-00022-of-00039.safetensors3.36 GB6cc8a86e1983084b087433e4f1602aad5dba71a332cebe166ebe4df1be09ec83
model-00023-of-00039.safetensors3.43 GBb5270967e6e4d247b2168b3ca7d57e4233008f338589303c69a5ffb37d7bc881
model-00024-of-00039.safetensors3.43 GB37f4475f0b2c12464d7fd82d59aacdb26c00e62352a21ff8a0c827bc913d5e2b
model-00025-of-00039.safetensors3.36 GBcbf24eda88d21b16f8c0dd49ce1eb56487151ea10fd44e2b7fb3fec970a99918
model-00026-of-00039.safetensors3.43 GBd61ee9fe9f640f6fae3df22f33cd28eeb3e7a06de05bee77548e5cc7cb4523ba
model-00027-of-00039.safetensors3.43 GB5b1205929d4f48fed78b02bd4e7af10599b8dd7b6a01c1d164b8b10e6d6b83a2
model-00028-of-00039.safetensors3.36 GB953a01901f4923ebb3e41ba9db213d73c48232d8f3e7645ada0c084c316c9a52
model-00029-of-00039.safetensors3.43 GB3bd8b8ebe2020a426c336541913f78fa5573cf619339463486c9c0f3245bb627
model-00030-of-00039.safetensors3.43 GBb3fa0fc14955291b6544d725e418593610178b4b176290858fade87569c2cbe3
model-00031-of-00039.safetensors3.36 GBda7b7f7d84aeb3a215ca8dc565c9d234d2689e7637fbff8ed8dd32be494ee8a6
model-00032-of-00039.safetensors3.43 GB962f8b829757b2fe7cb5b022cf78aa1b3f8b596f45fd39761c3472658916b9de
model-00033-of-00039.safetensors3.43 GB7754b12addbde168d9cbc995817909f7040784bb15522f567da96919e47d1119
model-00034-of-00039.safetensors3.36 GB416233ba769b8e23fe70a775a524064e07d1e2693feec08c6339aed4124c07ad
model-00035-of-00039.safetensors3.43 GBea3a546e8e3dcfd9c5dfe5c3d51fd9c1d9caa28ff631b03178bd914e1be4fbd5
model-00036-of-00039.safetensors3.43 GBede0cc863b54ed0cb533115a039c819aeb06e343f33e25caf814a20ad29bc569
model-00037-of-00039.safetensors3.36 GB1562aeb9fd4afae0e1d25468b8ac657132d4b61a22f990bf984ae44bd86b8370
model-00038-of-00039.safetensors3.43 GB9842675beca83181bfb0de49e3cbf6f4b29da8eb38f480d0d67c74bc7a885036
model-00039-of-00039.safetensors3.43 GBeee0de0658b97b076e5beefd2d2836238ab46c081fa7ce4e3e6f35f9ccdbc9e9
model.safetensors.index.json41.0 KB97804308c997e23e1e9c19726863122e48bf9a2a6de3a594435e69cfd0fcc3e8
tokenizer.json17.2 MB881c9c36c359e1617afef6f7583403567931b7b4f43f6552d2b2155a131650a2
tokenizer_config.json5.61 KBd188c2d2029544f54aa1be07d7621875535f0680ec432708d05bacabab1df623
VersionsEach time the pinned upstream commit or file list changed in this repo. Newest first.
Upstream commitDateContents
0ef753164b pinned now2026-11-15 04:20 UTC47 files · 134 GB

Use it

$ fw get laion/sft-repro-thinking-step630-nemotron-terminal-step1888
# manifest sha256 28dee073d4a4… · 134 GB · 47 files
$ fw verify ./sft-repro-thinking-step630-nemotron-terminal-step1888 laion/sft-repro-thinking-step630-nemotron-terminal-step1888
fw.pyChecks the signature and every sha256, tries each source in order, resumes.
$ python3 fw.py get https://freeweights.allternit.com/manifests/laion/sft-repro-thinking-step630-nemotron-terminal-step1888.json --dir ./sft-repro-thinking-step630-nemotron-terminal-step1888 $ python3 fw.py verify ./sft-repro-thinking-step630-nemotron-terminal-step1888 ./laion__sft-repro-thinking-step630-nemotron-terminal-step1888.json
hfThe Hugging Face CLI with one setting. Same repo id.
$ HF_ENDPOINT=https://freeweights.allternit.com hf download laion/sft-repro-thinking-step630-nemotron-terminal-step1888 --revision 0ef753164bb3004f23b5dca904f27e40710ce48e --local-dir ./sft-repro-thinking-step630-nemotron-terminal-step1888
transformersLoad the weights in Python.
import os os.environ["HF_ENDPOINT"] = "https://freeweights.allternit.com" from transformers import AutoModelForCausalLM, AutoTokenizer tok = AutoTokenizer.from_pretrained("laion/sft-repro-thinking-step630-nemotron-terminal-step1888", revision="0ef753164bb3004f23b5dca904f27e40710ce48e") model = AutoModelForCausalLM.from_pretrained("laion/sft-repro-thinking-step630-nemotron-terminal-step1888", revision="0ef753164bb3004f23b5dca904f27e40710ce48e", device_map="auto")
vLLMServe an OpenAI-compatible API.
$ HF_ENDPOINT=https://freeweights.allternit.com vllm serve laion/sft-repro-thinking-step630-nemotron-terminal-step1888 --revision 0ef753164bb3004f23b5dca904f27e40710ce48e

License

Name
Apache License 2.0
Id
Apache-2.0
Class
Permissive
Redistribution
yes
Listing
Hosted (signed)
Decided
License as declared by the publisher. No person read it. 2026-11-15
Text in
README.md
Text sha256
0271d5a32ba074958b15cfd425ae571918344b76599a64d477c7650111b46bc1
Terms gate
No extra terms beyond the license.

Read the license text

Licenses are checked against the license table. A custom license passes only after a person reads it.

Provenance

From: laion <huggingface.co/laion>Newsgroups: alt.freeweights.nemotronSubject: sft-repro-thinking-step630-nemotron-terminal-step1888Date: 2026-11-15Commit: 0ef7531Signed: 2DC28CC7B26A283B ✓
Upstream
laion/sft-repro-thinking-step630-nemotron-terminal-step1888
Pinned commit
0ef753164bb3004f23b5dca904f27e40710ce48e
Manifest
sft-repro-thinking-step630-nemotron-terminal-step1888.json · .minisig
Manifest sha256
28dee073d4a4034018257bd4f75c40e22f6e62d62cbd00b69c1fcb113971eba0
Signature
minisign, Ed25519, key 2DC28CC7B26A283B
Created
2026-11-15 04:20 UTC
Generator
freeweights-manifest/0.1.0
Takedowns
None.

How to verify

$ curl -O https://freeweights.allternit.com/manifests/laion/sft-repro-thinking-step630-nemotron-terminal-step1888.json $ curl -O https://freeweights.allternit.com/manifests/laion/sft-repro-thinking-step630-nemotron-terminal-step1888.json.minisig $ minisign -Vm sft-repro-thinking-step630-nemotron-terminal-step1888.json -P RWQ7KGqyx4zCLXtUFOqvfv8L3g+8fpKJ1FsuxPyOYMF+4KufbUxKq1Oq $ shasum -a 256 <file> # compare with the Files tab

A manifest whose signature fails must be rejected. A file whose size or sha256 does not match must be discarded. Verify a download

Model tree

No related models in this catalog.Nothing here lists this model as its parent, and it names no parent.
Get it
Download
$ HF_ENDPOINT=https://freeweights.allternit.com hf download laion/sft-repro-thinking-step630-nemotron-terminal-step1888
Open in Freeweights Studio

Pull, verify and run it in the Allternit app on Desktop, web or phone.

Signed manifest
minisign, key 2DC28CC7B26A283B. sft-repro-thinking-step630-nemotron-terminal-step1888.json. Provenance. Takedown history: none.
Source
huggingface.co/laion/sft-repro-thinking-step630-nemotron-terminal-step1888

Related models

All of alt.freeweights.nemotron →
From: laion <huggingface.co/laion>Subject: a3-rl-laion_nemotron-gym-agent-calendar-80-8B
From: ServiceNow-AI <huggingface.co/ServiceNow-AI>Subject: Apriel-Nemotron-15b-Thinker
From: lmstudio-community <huggingface.co/lmstudio-community>Subject: Apriel-Nemotron-15b-Thinker-GGUF
From: nvidia <huggingface.co/nvidia>Subject: Nemotron-3-Diarization
From: nvidia <huggingface.co/nvidia>Subject: nemotron-3.5-asr-streaming-0.6b
From: bartowski <huggingface.co/bartowski>Subject: nvidia_OpenCodeReasoning-Nemotron-14B-GGUF