mlx-community · llama · based on meta-llama/Llama-3.1-8B-Instruct
Llama-3.1-8B-Instruct-4bit
Llama-3.1-8B-Instruct-4bit by mlx-community: a 8B-parameter dense open-weight model under the llama3.1 license with tool calling.
1,046,370 downloads · 5 likes · Model card · synced 2026-09-19
Specifications
- Parameters
- 8B
- Architecture
- dense · llama
- Layers
- 32
- Hidden size
- 4,096
- KV heads · head dim
- 8 · 128
- Vocabulary
- 128,256
- Native dtype
- bfloat16
- Context window
- 131,072 tokens
- Released
- 2025-02-15
- Last modified on Hub
- 2025-02-15
Features & licensing
- License
- llama3.1
- Commercial use
- Yes
- Modalities
- text
- Tool / function calling
- Yes
- Reasoning mode
- No
- Languages
- EN, DE, FR, IT, PT, HI, ES, TH
- Quantised variants on Hub
- —
- Pipeline
- text-generation
Task fit
Editorial scores (0–100) used by the recommendation engine.
- Chat70
- RAG / Q&A70
- Code70
- Summarisation70
- Extraction70
- Agents70
Hardware to run Llama-3.1-8B-Instruct-4bit
Weights need about 18 GB at FP16, 9 GB at INT8 and 5 GB at INT4. The KV cache adds roughly 131 MB per 1,000 tokens per request. For a reference workload of 5 requests per second with 1,500 input and 300 output tokens, total GPU memory is around 7.8 GB, which fits on 1 × NVIDIA A100 (80 GB).
| Configuration | VRAM | Utilisation | Est. first token | Cloud / month |
|---|---|---|---|---|
| 1 × NVIDIA L4 (24 GB) | 24 GB | 32 % | ~0.5 s | $584 |
| 1 × NVIDIA L40S (48 GB) | 48 GB | 16 % | ~0.2 s | $1,387 |
| 1 × NVIDIA A100 (80 GB)recommended | 80 GB | 10 % | ~0.1 s | $2,336 |
| 1 × NVIDIA H100 (80 GB) | 80 GB | 10 % | ~0.0 s | $3,285 |
Estimates only. Run an assessment for your own traffic.
Auto-imported from Hugging Face. Architecture, license and feature data are synced from the model repository; the task-fit scores below are provisional defaults until a curator reviews them.