Gemmaverse: Open Gemma Variant Constellation & Hardware Benchmark

Ecosystem Constellation 8 Nodes

Base Checkpoint Community Adapter

Click nodes to inspect ancestry & benchmark gains

AstroGemma-Spectral-9B Space Science

Fine-tuned on high-resolution JWST and ESA spectroscopy telemetry for astrophysical classification.

Lineage: Gemma-2-9B Gain: +14.2% on SpaceQA

Quick Scenario Presets

Deployment Hardware Profiler

✓ Target VRAM OK

Base Architecture

Community Fine-Tune

Quantization Scheme

Target Device Hardware

Context Length: 8192 tokens

Concurrent Batch: 1

VRAM Allocation Gauge 6.45 GB / 24.0 GB

Model Weights (5.40 GB) Attention KV-Cache (1.05 GB)

Total VRAM Required 6.45 GB

Est. Generation Rate 42.8 tok/s

Domain Gain +14.2% on SpaceQA

Generated Deployment CLI Command (llama.cpp / vLLM)

llama-cli -m astrogemma-9b-q4_k_m.gguf -c 8192 -ngl 99 -b 1 --temp 0.2