pfnet/plamo-3-translate-31b
archived · low-signalmodel
PLaMo 3 Translate 31B is a translation-specialized LLM developed by Preferred Networks (PFN) that delivers translation quality. Dense, 32B params. Context: 256K tokens (config). License: plamo-community-license. Vendor-reported: "Even the 2B model outperforms the translation-specialized LLMs we compared against on every metric in the figure below."
- Vendor-reported: Even the 2B model outperforms the translation-specialized LLMs we compared against on every metric in the figure below.
- Org
- pfnet
- Released
- 2026-10-26
- Status
- archived · low-signal
- License
- plamo-community-license (restricted)
- Parameters
- 31.8B
- Architecture
- dense · Plamo3ForCausalLM
- Context
- 262,144 tokens (config.json)
- Task
- translation · text
- Signal
- 30 HF likes · 851 downloads
- Archived
- 2026-11-09 (low-signal: aged out below threshold; stats as last seen)
Runs on
estimate from params; need = weights + FP16 KV cache for 8K tokens + 1GB. A sizing calculation, not tested.
- Q4 ~22.2 GB (estimated size): 24GB GPU (3090 / 4090), 32GB GPU (5090), 24GB GPU (7900 XTX), Mac 36GB, Mac 48GB, CPU only, 32GB RAM (slow)
- Q8 ~36.8 GB (estimated size): 48GB GPU (RTX 6000 Ada / A6000), Mac 64GB (M1 Max), CPU only, 64GB RAM (slow)
- FP16 ~66.6 GB (estimated size): 96GB GPU (RTX Pro 6000 Blackwell), 1× DGX Spark (128GB), 2× DGX Spark (256GB), Strix Halo 128GB unified (~96GB usable), Mac 96GB, Mac 128GB, Mac 192GB, Mac 256GB, Mac 512GB, CPU only, 128GB RAM (slow), CPU only, 256GB RAM (slow)
Links
First published here 2026-10-26 · updated 2026-11-15 · markdown · JSON