Model

Llama 2 70B

OPEN

Meta · Llama 2 family · released Jul 18, 2023

First open-weights LLM licensed for commercial use at flagship scale; Llama 2's dense 70B base model with GQA for inference efficiency.

ReasoningCodingVisionFunction callingTool useAgentic

365.0

Elo · rank #341

Parameters

69B

Active params

69B (dense)

Context

4K tokens

Architecture

Dense decoder-only Transformer with Grouped-Query Attention (GQA)

License

Llama 2 Community License Agreement

Languages

—

API price (in/out)

No hosted API

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

AGIEvalReasoning54.2%#10

best: OLMo 3-Think 32B · 88.2%

ARC-ChallengeReasoning57.4%#100

best: Llama 3.1 405B · 96.9%

ARC-EasyReasoning80.2%#25

best: Phi-3-medium (14B) · 97.7%

BIG-Bench HardReasoning51.2%#98

best: ERNIE 4.5 300B-A47B · 94.3%

GSM8KMath56.8%#121

best: Llama 3.1 405B · 96.8%

HellaSwagReasoning85.3%#38

best: Claude 3 Opus · 95.4%

HumanEvalCoding29.9%#154

best: Claude Opus 4.5 · 99.4%

MBPPCoding45.0%#78

best: Llama-3.3-Nemotron-Super-49B v1 (Reasoning On) · 91.3%

MMLUKnowledge68.9%#154

best: OpenAI o3 · 92.9%

OpenBookQAReasoning60.2%#13

best: Claude 1 · 90.8%

PIQAReasoning82.8%#20

best: GPT-4o mini · 93.1%

Social IQaReasoning50.7%#23

best: Apple DCLM-Baseline 7B · 82.9%

TriviaQAKnowledge87.6%#4

best: Sarvam-1 (2B) · 90.6%

TruthfulQAKnowledge50.2%#57

best: Phi-3.5-MoE (16x3.8B, 6.6B active) · 77.5%

WinoGrandeReasoning80.2%#37

best: PaLM 2 · 90.9%

Run it locally

VRAM @ Q4

38 GB

VRAM @ FP16

140 GB

Fits on (Q4)

M3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

Q4_K_M weights (~38-41GB) exceed the RTX 4090's 24GB VRAM, forcing CPU offload; no reliably sourced single-GPU 4090 tokens/sec figure was found specifically for this model.

Quantizations

GGUF · GPTQ · AWQ · GGML

Llama 2 family

Elo progression across releases

Llama 2 13BJul 202369.7 Llama 2 13B ChatJul 2023105.1 Llama 2 70BJul 2023365.0 Llama 2 70B ChatJul 2023363.8 Llama 2 7BJul 2023-121.8 Llama 2 7B ChatJul 2023-10.6

Resources

HFmeta-llama/Llama-2-70b GHmeta-llama/llama DOCSai.meta.com/llama APIAvailable via Meta API

API price weights · each benchmark row carries its own source badge (see methodology)