What are the hardware requirements for Reka Flash 3?

Reka Flash 3 requires approximately 41.7 GB of VRAM and supports a context window of 32K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Reka Flash 3 and how large is it?

Reka Flash 3 is developed by RekaAI, a model with 20.9b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate Reka Flash 3?

Reka Flash 3 is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Reka Flash 3 by RekaAI — VRAM 41.7GB, 32K context

Name: Reka Flash 3
Author: RekaAI

Reka Flash 3 is an open-source language model by RekaAI. Features: 20.9b LLM, VRAM: 41.7GB, Context: 32K, License: apache-2.0, LLM Explorer Score: 0.19.

Llama Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/RekaAI/reka-flash-3

Reka Flash 3 Parameters and Internals

LLM Name	Reka Flash 3
Repository 🤗	https://huggingface.co/RekaAI/reka-flash-3
Model Size	20.9b
Required VRAM	41.7 GB
Updated	2026-07-03
Maintainer	RekaAI
Model Type	llama
Model Files	9.8 GB: 1-of-5 9.8 GB: 2-of-5 9.8 GB: 3-of-5 9.8 GB: 4-of-5 2.5 GB: 5-of-5
Model Architecture	LlamaForCausalLM
License	apache-2.0
Context Length	32768
Model Max Length	32768
Tokenizer Class	GPT2Tokenizer
Vocabulary Size	100352
Torch Data Type	bfloat16

Rank the Reka Flash 3 capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to Reka Flash 3

Best Alternatives	Context / RAM	Downloads	Likes
Reka Flash 3.1	96K / 41.7 GB	90	101

Note: green Score (e.g. "73.2") means that the model is better than RekaAI/reka-flash-3.