Reka Flash

Reka Flash

Reka Flash

by

Reka

Model ID:

reka-flash

Use This Model

Model Overview

FAQs

Related Links

Overview

Reka Flash is a 21B parameter dense model that punches well above its weight class, often outperforming much larger models like Llama 2 70B and Gemini Pro. It achieves impressive scores on benchmarks like MMLU, GSM8K, and HumanEval despite its relatively small size. Flash also demonstrates strong multimodal capabilities, performing competitively on visual question answering tasks. In human evaluations, it ranks highly for multimodal chat. Flash is an excellent choice for developers who need near-frontier model performance but have more limited computational resources. Its combination of strong language understanding, reasoning, and multimodal abilities in a smaller package makes it well-suited for a wide range of applications where balancing capability and efficiency is important.

Specializations

  • Exceptional performance for its size (21B parameters.

  • Strong multimodal abilities, especially in visual question answering.

  • Outperforms many larger models on standard benchmarks.

  • Efficient balance of capability and computational requirements.

  • Excels in coding tasks (e.g., high HumanEval scores).

  • Competitive performance in multilingual tasks.

  • Strong in video understanding tasks.

Integration Guide (Javascript)

To use this model through Portkey, follow these steps:

1. Install Portkey SDK:

npm install --save portkey-ai

2. Set up client with Portkey:

// Import and initialize Portkey

import Portkey from 'portkey-ai'

const portkey = new Portkey({

  apiKey: "PORTKEY_API_KEY", // Replace with your Portkey API key

  virtualKey: "VIRTUAL_KEY" // Your Reka AI Virtual Key created in Portkey

})

3. Make a request:

const chatCompletion = await portkey.chat.completions.create({

    messages: [{ role: 'user', content: 'Say this is a test' }],

    model: 'reka-flash',

});

console.log(chatCompletion.choices);

Live API Statistics and Performace Metrics

LMSys Arena Elo Score:

1200

Arena Score

LMSys Leaderboard Rank:

36

Arena Rank

Evaluates LLMs through a combination of user feedback and automated scoring systems.

Model Specifications

Release Date:

15/4/2024

Max. Context Tokens:

128K

Model Size

21B

Knowledge Cut-Off Date:

November 2023

MMLU:

75.9

%

License:

Proprietary

Technical Report/Model Card:

LMSys Elo Score

1200

Pricing

$/Million Input Tokens

$

0.2

$/Million Output Tokens

$

0.8

Live updates via Portkey Pricing API. Coming Soon...

© 2024 Portkey, Inc. All rights reserved