ModelListedExternal

Jupiter G 8B

Sourcehuggingface.co/locailabs/Jupiter-G-8B/tree/4eb096796a8f1fa7390bfdfa14a8aad85abafab1

KindText modelBase modelgoogle/gemma-4-E4B-itVersionmodel

Model source

Kind: Text model
File size: 14.9 GB
Base model: google/gemma-4-E4B-it
Version: vmodel
Updated: Sep 15, 2026
Downloads: 74
Source: Hugging Face

Source description

Jupiter

Jupiter-G-8B

Jupiter-G-8B is a post-trained variant of Google Gemma 4 E4B IT, developed by Locai Labs. The G denotes the Gemma base. Jupiter-G-8B improves instruction following (+1.7 IFEval, +1.0 IFBench) and coding/agentic capability (+3.2 LCB pass@1) while preserving the base model's reasoning and knowledge through our Forget-Me-NotTM framework. This model was trained on 1 H200 GPU using 100% renewable energy.

Benchmarks

We evaluate Jupiter-G-8B against gemma-4-E4B-it.

Benchmark	Jupiter-G-8B	gemma-4-E4B-it
IFEval (prompt strict)	89.3	87.6
IFBench (prompt strict)	35.4	34.4
AgentHarm harm rate	12.0	22.3
MMLU Redux	82.0	83.4
LiveCodeBench v6	55.2	52.0

IFEval and IFBench both reported with prompt strict accuracy. LiveCodeBench v6 reported with pass@1.

Training

Post-Training Data

Jupiter-G-8B is fine-tuned on a curated mixture of five datasets:

Dataset	Domain	N
Self-cognition (non-reasoning)	Identity	~full
UltraChat (reasoning + non-reasoning)	Reasoning / Replay	12,500
Nemotron terminal trajectories (reasoning)	Terminal / Agentic	20,000
Nemotron competitive programming (non-reasoning)	Coding	20,000

Training Configuration

Method	LoRA (rank 16, alpha 32)
Target Modules	All linear layers
Epochs	2
Optimiser	AdamW (fused)
Learning rate	2e-4 (cosine decay, 5% warmup)
Weight decay	0.001
Max grad norm	1.0
Batch size	64 (global: 8 local x 8 accumulation)
Sequence length	2,048
Precision	BF16
Gradient checkpointing	Enabled
Loss	Assistant-only
Kernel	Liger
Attention	Eager

Key Techniques

Forget-Me-Not: Synthetic replay data generated by the unmodified base model on UltraChat prompts, preserving existing capabilities during domain-specific fine-tuning.
Agentic/terminal training: Curated terminal trajectories from NVIDIA's Nemotron-Terminal-Corpus.
Competitive programming: Exercism-derived programming problems from Nemotron corpora to strengthen code generation.

Citation

@misc{locailabs2026jupiterg,
  title   = {Jupiter-G-8B},
  author  = {George Drayson},
  year    = {2026},
  url     = {https://huggingface.co/locailabs/Jupiter-G-8B}
}

Acknowledgements

Jupiter-G-8B builds on Google Gemma 4. Terminal and programming data are sourced from NVIDIA's Nemotron corpora.

Sources

1 source

Hugging FacePrimaryVersion 4eb096796a8f1fa7390bfdfa14a8aad85abafab1

huggingface.co/locailabs/Jupiter-G-8B/tree/4eb096796a8f1fa7390bfdfa14a8aad85abafab1

Verified Sep 15

Model artifacts

1 artifact

model.safetensors

safetensors · 14.9 GB · SHA-256 5933dd4c6624…ae4b · Hugging Face

Download

Source excerpts

2 excerpts

ReadmeSep 15huggingface.co/locailabs/Jupiter-G-8B

--- license: apache-2.0 library_name: transformers pipeline_tag: any-to-any base_model: - google/gemma-4-E4B-it tags: - locai - jupiter - gemma - post-training --- ![Jupiter](jupiter.png) # Jupiter-G-8B Jupiter-G-8B is a post-trained variant of [Google Gemma 4 E4B IT](https://huggingface.co/google/gemma-4-E4B-it), developed by [Locai Labs](https://locailabs.com/). The G denotes the Gemma base. Jupiter-G-8B improves instruction following (+1.7 IFEval, +1.0 IFBench) and coding/agentic capability (+3.2 LCB pass@1) while preserving the base model's reasoning and knowledge through our Forget-Me-Not™ framework. This model was trained on 1 H200 GPU using 100% renewable energy. ## Benchmarks We evaluate Jupiter-G-8B against [gemma-4-E4B-it](https://huggingface.co/google/gemma-4-E4B-it). | Benchmark | Jupiter-G-8B | gemma-4-E4B-it | | --- | --- | --- | | IFEval (prompt strict) | **89.3** | 87.6 | | IFBench (prompt strict) | **35.4** | 34.4 | | AgentHarm harm rate | **12.0** | 22.3 | | MMLU Redux | 82.0 | **83.4** | | LiveCodeBench v6 | **55.2** | 52.0 | IFEval and IFBench both reported with prompt strict accuracy. LiveCodeBench v6 reported with pass@1. ## Training ### Post-Training Data Jupiter-G-8B is fine-tuned on a curated mixture of five datasets: | Dataset | Domain | N | | --- | --- | --- | | Self-cognition (non-reasoning) | Identity | ~full | | UltraChat (reasoning + non-reasoning) | Reasoning / Replay | 12,500 | | Nemotron terminal trajectories (reasoning) | Terminal / Agentic | 20,000 | | Nemotron competitive programming (non-reasoning) | Coding | 20,000 | ### Training Configuration | Method | LoRA (rank 16, alpha 32) | | --- | --- | | Target Modules | All linear layers | | Epochs | 2 | | Optimiser | AdamW (fused) | | Learning rate | 2e-4 (cosine decay, 5% warmup) | | Weight decay | 0.001 | | Max grad norm | 1.0 | | Batch size | 64 (global: 8 local x 8 accumulation) | | Sequence length | 2,048 | | Precision | BF16 | | Gradient checkpointing | Enabled | | Loss | Assistant-only | | Kernel | Liger | | Attention | Eager | ### Key Techniques - **Forget-Me-Not:** Synthetic replay data generated by the unmodified base model on UltraChat prompts, preserving existing capabilities during domain-specific fine-tuning. - **Agentic/terminal training:** Curated terminal trajectories from NVIDIA's Nemotron-Terminal-Corpus. - **Competitive programming:** Exe...

NoteSep 15huggingface.co/locailabs/Jupiter-G-8B

Source context: 74 downloads · 5 likes · Pipeline any-to-any · Library transformers · Repo locailabs/Jupiter-G-8B