// Ai2 OLMo / Molmo · Text LLM
Llama-3.1-Tulu-3-405B: VRAM & local deployment
405.9B parameters. About 275 GB at 4-bit and 898 GB at BF16 with one 8K-token request.
released 2025-01-09licence: Commercial with conditionsmemory figures are estimates
01Specs
| Publisher | Allen Institute for AI · allenai |
|---|---|
| Family / series | Ai2 OLMo / Molmo · Llama-3.1-Tulu-3 |
| Type | Text LLM (text-generation) |
| Parameters | 405.9B Source: counted from official safetensors weights |
| Context length | 128K tokens (131,072)max_position_embeddings in config.json |
| Architecture | 126 layers · GQA |
| Released weights | BF16 · 812 GBdownload |
| Licence | Llama 3.1 Community · Commercial with conditionsCommercial use allowed under the publisher’s own licence terms (user caps, acceptable-use policy, attribution or region limits). Read it before shipping. |
| Released | 2025-01-09 (Hugging Face repo created) |
02How much VRAM it needs (estimate)
Total = weights + 10% + 0.5GB overhead + KV cache for one request (FP16 KV). More concurrent users multiply the KV column.
| Precision | Weights | Total @ 8K | Total @ 32K | Total @ 128K | Smallest GPU | Mac |
|---|---|---|---|---|---|---|
| BF16 / FP16 (original) | 812 GB | 898 GB | 910 GB | 961 GB | 8 × H200 (1,128GB) | — |
| FP8 / INT8 / Q8_0 | 431 GB | 479 GB | 492 GB | 542 GB | 8 × H100 (640GB) | — |
| 4-bit (Q4_K_M / AWQ / GPTQ-Int4) | 246 GB | 275 GB | 288 GB | 339 GB | 4 × RTX PRO 6000 (384GB) | Mac 512GB |
KV cache per request: 8K 4.2 GB · 32K 16.9 GB · 128K 67.7 GB
Bits per weight: BF16 16, FP8 8.5, 4-bit 4.85. “Fits” = total ≤ 90% of memory; Mac GPU can use ~2/3 of unified memory (≤36GB) or ~3/4 (larger). For MoE, all experts must be in memory, so size by total parameters. Method →
03How to run it
Only runtimes the publisher or the runtime’s official library names for this model.
- vLLM — mentioned in the official model card
- Transformers — mentioned in the official model card
04Official repositories & quantised builds
1 official repo(s) from allenai.
05Other Ai2 OLMo / Molmo models
All 31 →| Model | Params | 4-bit total | Released |
|---|---|---|---|
| Llama-3.1-Tulu-3-8B · same series | 8.03B | 6.9 GB | 2024-11-20 |
| Molmo2-ER | 4.85B | 4.9 GB | 2026-05-04 |
| MolmoWeb-4B | 4.85B | 4.9 GB | 2026-03-20 |
| MolmoPoint-8B | 8.68B | 7.5 GB | 2026-03-16 |
| Olmo-Hybrid-7B | 7.43B | 6.5 GB | 2026-01-28 |
| Molmo2-4B | 4.85B | 4.9 GB | 2025-12-14 |
| Molmo2-8B | 8.66B | 7.5 GB | 2025-12-14 |
| Molmo2-O-7B | 7.76B | 10.0 GB | 2025-12-14 |
| Olmo-3.1-32B-Instruct | 32.2B | 23.3 GB | 2025-12-10 |
06Sources
Data checked 2025-02-10. Parameters, context and licence are copied from the official repository; memory figures are our estimates.