Model

Llama 2 7B

OPEN

Meta · Llama 2 family · released Jul 18, 2023

Compact Llama 2 base model with 2T-token pretraining and a 4K context window, commercially licensed unlike its LLaMA 1 predecessor.

ReasoningCodingVisionFunction callingTool useAgentic

-121.8

Elo · rank #414

Parameters

6.7B

Active params

6.7B (dense)

Context

4K tokens

Architecture

Dense decoder-only Transformer

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

AGIEvalReasoning29.3%#38

best: OLMo 3-Think 32B · 88.2%

ARC-ChallengeReasoning45.9%#133

best: Llama 3.1 405B · 96.9%

ARC-EasyReasoning75.2%#36

best: Phi-3-medium (14B) · 97.7%

BIG-Bench HardReasoning32.6%#124

best: ERNIE 4.5 300B-A47B · 94.3%

GSM8KMath14.6%#172

best: Llama 3.1 405B · 96.8%

HellaSwagReasoning77.2%#97

best: Claude 3 Opus · 95.4%

HumanEvalCoding12.8%#180

best: Claude Opus 4.5 · 99.4%

MBPPCoding20.8%#106

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

MMLU (EU-21 languages)Knowledge36.4%#15

best: Llama 3.1 70B · 77.1%

MMLUKnowledge45.3%#245

best: OpenAI o3 · 92.9%

PIQAReasoning78.8%#51

best: GPT-4o mini · 93.1%

Social IQaReasoning48.3%#30

best: Apple DCLM-Baseline 7B · 82.9%

TriviaQAKnowledge72.1%#21

best: Sarvam-1 (2B) · 90.6%

TruthfulQAKnowledge33.3%#100

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

WinoGrandeReasoning69.2%#106

best: PaLM 2 · 90.9%

Run it locally

VRAM @ Q4

6.6 GB

VRAM @ FP16

13.5 GB

Fits on (Q4)

RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

No verifiably-sourced Llama-2-7B-base-specific RTX 4090 llama.cpp benchmark was found; only Llama 2 13B (chat) had a confirmed on-topic measurement.

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-2-7b GHmeta-llama/llama DOCSai.meta.com/llama APIAvailable via Meta API

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