MODELS
Llama 3 8B
Meta's mass-deployable 8B open weights — runs anywhere from a laptop up.
Specs
- Context window
- 8,192
- Max output
- 4,096
- Modalities
- text
- Tool use
- —
- Vision
- —
- Streaming
- ✓
- License
- llama-3-community
- Released
- 2024-04-18
Pricing
Llama 3 8B (April 2024) is Meta's small open-weights model — the practical default for self-hosted LLM deployments because it fits on a single consumer GPU (16GB), a Mac with 16GB unified memory, or even quantised on Raspberry Pi 5. 8K context, no native tool use or vision. Llama 3 Community Licence (commercial-friendly with a 700M MAU clause). Successor: Llama 3.1 8B with 128K context.
Editor's verdict
Pick this when you need on-device or air-gapped LLM and don't want to pay per-token. Quality is well below Sonnet 4.6 or GPT-5 — don't expect frontier reasoning — but for entity extraction, classification, RAG synthesis, draft writing it's plenty. Strong Chinese underwhelms; for Mandarin/Cantonese workloads consider Qwen 2.5 7B or Yi instead.
Reviews
No reviews yet. Be the first.
Last updated: 2026-04-29