Reka AI/Dense

RReka Flash 3

Name: Reka Flash 3
Author: Reka AI

Reka Flash 3 is a 21B general-purpose reasoning model that was trained from scratch. It was trained in synthetic and public datasets for supervised finetuning, followed by RLOO with model-based and rule-based rewards.

chatreasoningcoding

21B

Parameters

32K

Context length

Benchmarks

Quantizations

Architecture

Dense

Released

2025-03-15

Layers

KV Heads

Head Dim

Family

other

Quantization Options

Quant	Bits	VRAM	Quality
Q3_K_M	4	11.0 GB	low
Q3_K_L	4.3	11.8 GB	moderate
IQ4_XS	4.46	12.2 GB	moderate
Q4_K_S	4.67	12.7 GB	moderate
Q4_K_M	4.89	13.3 GB	good
Q5_K_S	5.57	15.1 GB	good
Q5_K_M	5.7	15.5 GB	good
Q6_K	6.56	17.7 GB	excellent
Q8_0	8.5	22.8 GB	lossless
FP16	16	42.5 GB	lossless

Select your GPU above to see speed estimates and compatibility for each quantization.

▸ READY TO RUN THIS?RENT BY THE HOUR

RENT A GPU AND RUN REKA FLASH 3 NOW

Rent on RunPod →Or Vast.ai →

Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.

Community Ratings

Loading ratings...

Benchmarks (11)

MATH-50089.3

MMLU-PRO65.0

GPQA Diamond52.9

LiveCodeBench43.5

AIME33.7

AA Math33.7

IFBench30.4

SciCode26.7

AA Intelligence9.5

AA Coding8.9

HLE5.1

Run this model

▸Easiest way to get started·Beginners

DOCS ↗

curl -fsSL https://ollama.com/install.sh | sh

$ollama run other:21b-q4_K_M

Tag may need adjustment — check ollama.com/library/other for available tags.

▸ SETUP GUIDE

Auto-setup with fitmyllm CLI

Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.

pip install fitmyllmthen run fitmyllmLearn more

Auto-detect GPULive tok/s in chatSpeed benchmarks9 inference engines

HuggingFace GGUF Downloads Build Hardware

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

NVIDIA RTX 5080

16 GB VRAM • 960 GB/s

NVIDIA

$999