Chat

Llama 3 Community LicenseOpen weightsUpdated June 2026Legacy baseline

Llama 3 70B

Meta open-weight model family still commonly used as a baseline for local AI stacks and app prototypes.

Meta · Llama

Editorial review

Reviewed byOpenSourcesAI EditorialLast updatedJune 2026SourcesExact model card

Model checkpoints, context windows, provider support, local runtime compatibility, and license terms can change quickly. Verify the exact model card before production or commercial use.

Best for

Builders who want a widely supported open-weight chat model with broad runtime compatibility.

Who should use it

  • Builders who want a widely supported open-weight chat model with broad runtime compatibility.
  • Builders who want local or self-hosted testing options.

Common workflows

  • General chat, instruction following, local app prototypes
  • chat workflows
  • local workflows
  • baseline workflows

Deployment and hardware notes

Quantized 70B-class models usually need high-memory GPUs or unified memory systems.

License and usage notes

Llama 3 Community License. Open weights. Verify the exact model card and license terms for the checkpoint or hosted provider you use.

Strengths

  • Open weights model option for Llama workflows.
  • Builders who want a widely supported open-weight chat model with broad runtime compatibility.
  • Commonly used in local workflows through quantized builds, but 70B-class models are best with high-memory GPUs or workstation/server hardware.
  • Tracked as Legacy baseline in the OpenSourcesAI model directory.

Limitations

  • License is not a standard open-source license; newer models may outperform it for coding and reasoning.
  • Quantized 70B-class models usually need high-memory GPUs or unified memory systems.
  • Context window and limits: 8,192 tokens, confirmed from the model's published config.
  • Verify the exact model card, provider docs, license, and serving support before production use.

Local workflow notes

Commonly used in local workflows through quantized builds, but 70B-class models are best with high-memory GPUs or workstation/server hardware.

Local runtimes: Ollama, LM Studio, llama.cpp, vLLM

Platforms: Windows, macOS, Linux, Workstations

Will Llama 3 70B run on your machine?

Llama 3 70B is 70B parameters and needs 43 GB of VRAM at Q4_K_M — 41.5 GB of weights plus 1.5 GB of runtime overhead for the inference server itself.

VRAM by quantization

QuantizationWeightsNeeds (with overhead)Quality
Q4_K_S39 GB40.5 GBgood
Q4_K_M41.5 GB43 GBgood
Q8_074 GB75.5 GBhigh
FP16140 GB141.5 GBreference

Fit on common hardware at Q4_K_M

HardwareMemory the model can useSystem RAMVerdict
CPU OnlyNone (CPU only)16 GBToo large
RTX 4060 Laptop8 GB16 GBToo large
RTX 3060 (12GB)12 GB32 GBToo large
RTX 4060 Ti (16GB)16 GB32 GBToo large
RTX 309024 GB64 GBToo large
Apple Silicon (Unified Memory) 36 GB27 GB of 36 GB36 GBToo large
RTX 509032 GB64 GBToo large
Apple Silicon (Unified Memory) 48 GB36 GB of 48 GB48 GBToo large
Apple Silicon (Unified Memory) 64 GB48 GB of 64 GB64 GBComfortable
Apple Silicon (Unified Memory) 96 GB72 GB of 96 GB96 GBComfortable
Apple Silicon (Unified Memory) 128 GB96 GB of 128 GB128 GBComfortable

Comfortable means VRAM clears the requirement by 2 GB or more. Tight means it covers the requirement with no margin. CPU offload means the model does not fit in VRAM but system RAM is at least 1.6× the weights, so it will run at reduced speed — expect roughly 1–5 tokens per second. Figures are weights plus a fixed runtime overhead and exclude KV-cache growth, which scales with context length.

Apple Silicon shares one pool of memory between the system and the GPU, so a model cannot use all of it. These rows apply the same 75% usable fraction the Compatibility Checker uses, which is why a 36 GB Mac is graded on less than 36 GB.

Check Llama 3 70B against your exact hardwareHow these verdicts are calculated

Need more hardware for Llama 3 70B? Open the PC Builder for the 70B tier →

VRAM fit by quantization level

Enter your GPU VRAM below to see which quantization of Llama 3 70B fits and get the Ollama run command.

Sources to verify

Exact model card

Related resources

Continue with model source notes, local tools, and implementation guides related to this model.

Llama family hubBest local LLM toolsPrivate local chatbot stackOllama + Open WebUI starter stackAI tools directory
Hardware48GBRuntimeOllama, llama.cpp, vLLM, LM StudioContext8,192 tokens, confirmed from the model's published configLast updated2026
Exact model card →

Model ecosystem connections

Use these next-step links to move from this profile into related tools, comparisons, guides, stacks, and curated shortlists.

Compatible hardware

Apple Silicon 64GB Unified Memory →Apple Silicon 96GB Unified Memory →Apple Silicon 128GB Unified Memory →Apple Silicon 192GB Unified Memory →Dual RTX 3090 48GB →Dual RTX 4090 48GB →Quad RTX 3090 96GB →

Recommended runtimes and tools

Ollama →LM Studio →llama.cpp →vLLM →

Setup and deployment

Check if this model fits your GPU →What can 48GB VRAM run? →What is quantization? GGUF, Q4, Q8 explained →What is VRAM and how much do you need? →

Related model pages

Llama family hub →Llama 3 8B Instruct →Llama 4 Maverick →Llama 4 Scout →Mistral 8x22B Instruct →Mistral Large 2 →Llama 3.3 70B →

Guides, stacks, and comparisons

Best local LLM tools →Private local chatbot stack →Ollama + Open WebUI starter stack →All model comparisons →

Related models

Llama 3 8B InstructLocal baseline comparisons and lightweight app prototypes.Llama 4 MaverickBuilders comparing current Llama-family models for assistant, multimodal, and reasoning-oriented workflows.Llama 4 ScoutTeams evaluating Llama-family models for multimodal assistant, long-context, and application workflows.Mistral 8x22B InstructMoE assistant evaluationMistral Large 2Large multilingual assistant evaluationLlama 3.3 70B70B open-weight model in the Llama family.

Ready to run this model locally?

Find a compatible interface in our Local AI Tools directory →