Model

RedPajama-INCITE-7B-Base

OPEN

Together · RedPajama-INCITE family · released Jun 6, 2023

Fully reproducible 7B trained on 1T RedPajama-1T tokens with 10 released checkpoints, Apache-2.0.

ReasoningCodingVisionFunction callingTool useAgentic

-299.7

Elo · rank #432

Parameters

6.9B

Active params

6.9B (dense)

Context

2K tokens

Architecture

GPT-NeoX / Pythia-style dense transformer

License

Apache-2.0

Languages

—

API price (in/out)

No hosted API

Modalities

text

Benchmark results

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

ARC-ChallengeReasoning39.3%#151

best: Llama 3.1 405B · 96.9%

ARC-EasyReasoning69.2%#51

best: Phi-3-medium (14B) · 97.7%

HellaSwagReasoning70.4%#123

best: Claude 3 Opus · 95.4%

LAMBADAReasoning71.4%#14

best: GPT-3 175B · 86.4%

MMLUKnowledge32.3%#266

best: OpenAI o3 · 92.9%

OpenBookQAReasoning50.4%#24

best: Claude 1 · 90.8%

PIQAReasoning77.4%#58

best: GPT-4o mini · 93.1%

TruthfulQAKnowledge20.5%#109

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

WinoGrandeReasoning64.2%#126

best: PaLM 2 · 90.9%

Run it locally

VRAM @ Q4

5 GB

VRAM @ FP16

14 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

Throughput data unavailable.

Quantizations

GGUF Q4

RedPajama-INCITE family

Elo progression across releases

RedPajama-INCITE-Base-3B-v1May 2023-470.3 RedPajama-INCITE-7B-BaseJun 2023-299.7 RedPajama-INCITE-7B-ChatJun 2023-382.3 RedPajama-INCITE-7B-InstructJun 2023-300.8

Resources

HFtogethercomputer/RedPajama-INCITE-7B-Base DOCStogether.ai/blog/redpajama-7b

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