MA
Meta AI
Language model · United States ·Jul 2024

Llama 3.1-70B

Language Open (Restricted)
70B
Parameters
20
Benchmarks

About Llama 3.1-70B

Llama 3.1-70B is an AI model developed by Meta AI, in the language category, released in 2024, made available as an open-weights model with 70B params.

On this page you'll find Llama 3.1-70B's full specifications, including 20 benchmark results. Review provider pricing and benchmark scores below, or compare Llama 3.1-70B head-to-head with other language models.

→ More models by Meta AI → All language models → AI benchmarks leaderboard

Links

Announcement ↗ Hugging Face ↗

Run Llama 3.1-70B locally — what to buy

Share this build ↗

Sized to what Llama 3.1-70B actually needs (~8k context at Q4), not to the biggest GPU: Good = cheapest that runs it, Better = best value, Best = most headroom. Speed is a hardware estimate; anything past ~120 tok/s is shown as instant.

Best
Apple M1 Ultra 64GB
Apple · ~$3,700
~16.7 tok/s
Q4 · fits full context · up to 8k ctx
Runs with: ExLlamaV2, KoboldCpp, LM Studio
See full estimate →

Software support

·ExLlamaV2 ·KoboldCpp ·LM Studio ·MLX ·Ollama ·SGLang ·TensorRT-LLM ·llama.cpp ·vLLM ?KTransformers ?text-generation-webui

✓ measured · · compatible · — not supported. Informational only — speed is hardware-based.

Consumer GPU
5–6.1 tok/s
2 options →
Apple
8.4–16.7 tok/s
13 options →
Datacenter GPU
17–120+ tok/s
13 options →

Community

Would you run Llama 3.1-70B again?
No votes yet — be the first.
Did it run? — community reports by hardware

No reports yet. Own this model on some hardware? Be the first to confirm it runs.

Discussion

No comments yet. Start the discussion.

Llama 3.1-70B benchmark scores▶
Compare on the benchmark table →
Benchmark Category Score Variant
HumanEval code 93.6% instruct-pass@10-cited-tulu3
HumanEval code 80.5% cited-telechat2
HumanEval code 80.5% instruct-cited-nvlm
HumanEval+ code 89.5% instruct-pass@10-cited-tulu3
MBPP code 86% cited-telechat2
AlpacaEval 2 general 33.4% instruct-lc-win-cited-tulu3
IFEval general 88% instruct-cited-tulu3
IFEval general 87.5% cited-telechat2
MMLU general 86% cited-telechat2
MMLU general 85.3% instruct-cited-tulu3
MMLU general 83.6% instruct-cited-nvlm
CMMLU knowledge 69.01 cited-telechat2
GSM8K math 95.1% cited-telechat2
GSM8K math 95.1% instruct-cited-nvlm
GSM8K math 93.7% instruct-cited-tulu3
MATH math 68% instruct-cited-nvlm
MATH math 56.4% instruct-cited-tulu3
BIG-Bench Hard reasoning 73.8% instruct-cited-tulu3
DROP reasoning 77% instruct-cited-tulu3
TruthfulQA safety 66.8% instruct-cited-tulu3

More models like Llama 3.1-70B

AI
DMind-1
 AI
COLLECTIVE-1
 AI
Cogito 1 series
 AI
Euler (Euler大模型)
· 2.5B
 AI
SmolLM-1.7B
· 1.7B
 AI
Llama-3-70B-Synthia-v3.5
· 70B

Frequently asked questions

What is Llama 3.1-70B? ▶

Llama 3.1-70B is an AI model developed by Meta AI, in the language category, released in 2024. It is tracked on GenAIList with its specifications, benchmark scores and provider pricing.

Who created Llama 3.1-70B? ▶

Llama 3.1-70B was developed by Meta AI and released in 2024.

Is Llama 3.1-70B open source or proprietary? ▶

Llama 3.1-70B ships as an open-weights model: you can download and self-host the weights, though the licence may place some restrictions on use.

How much does Llama 3.1-70B cost? ▶

Pricing for Llama 3.1-70B depends on the provider. See the providers table on this page for the latest API rates.

How does Llama 3.1-70B perform on benchmarks? ▶

Llama 3.1-70B is benchmarked across 20 evaluations on GenAIList, including HumanEval (93.6%). See the full benchmark table below and compare it with other models.

No videos yet. Know a good one? Add it below.

Know a great video about Llama 3.1-70B?