Rekaai · all models ›

RReka Flash 3OPEN

Reka Flash 3 is a general-purpose, instruction-tuned large language model with 21 billion parameters, developed by Reka.

66KContext window · tokens
$0.1Input price · per M tokens
$0.2Output price · per M tokens
RekaaiProvider

Prices update automatically — checked daily against provider list prices.

See model comparisons → Compare all model prices

Benchmarks & performance

Independent benchmark scores for Reka Flash 3, measured by Artificial Analysis. Higher is better.

Intelligence index3.7
Math index33.7
GPQA52.9%
MMLU-Pro66.9%
Humanity's Last Exam4.4%
Long Context Reasoning0%
LiveCodeBench43.5%
SciCode26.7%
MATH-50089.3%
AIME51%
AIME 202533.7%
IFBench30.4%
τ²-Bench0%
Terminal-Bench Hard0%
⚡ Speed88.2 tokens/sec
⏱ Latency1.33s to first token
💰 Blended price$0.35 / 1M tokens
📈 Value10.6 intelligence points per $
vs. models measured heretop 97%
Scores higher than 3% of the 273 models measured by Artificial Analysis and tracked here.
Models at this level cost $0.5 per 1M tokens (median of 11) — this one costs $0.13.
Cheaper and better on this index: GLM 5.3 Flash · DeepSeek V4 Flash 0731 · DeepSeek V4 Flash 0423 and 20 more
Benchmark data by Artificial Analysis

About this model

Reka Flash 3 is an open-weight AI model by Rekaai. You can download and self-host it for free; the prices below are hosted-API list prices, tracked daily, for when you prefer convenience over self-hosting.

Frequently asked questions

What is Reka Flash 3?

Reka Flash 3 is an AI language model from Rekaai. It is open-weight: you can download it and run it on your own hardware, for free. It scores 3.7 on the Artificial Analysis intelligence index.

Is Reka Flash 3 free?

The weights are free and open — you can self-host Reka Flash 3 and pay nothing per token. If you prefer a hosted API, list prices are $0.1 per million input tokens and $0.2 per million output tokens.

What is Reka Flash 3 good at?

Independent benchmarks from Artificial Analysis give it GPQA 52.9%, MMLU-Pro 66.9%, Humanity's Last Exam 4.4%, Long Context Reasoning 0%, LiveCodeBench 43.5%, SciCode 26.7%, MATH-500 89.3%, AIME 51%, AIME 2025 33.7%, IFBench 30.4%, τ²-Bench 0%, Terminal-Bench Hard 0%. It is particularly used for mathematical reasoning.

How fast is Reka Flash 3?

It generates about 88.2 tokens per second, with a median 1.33s delay before the first token. Measured independently by Artificial Analysis.

Can I self-host Reka Flash 3?

Yes. Reka Flash 3 has open weights, so you can download it and run it on your own GPU or server with tools like Ollama, vLLM or llama.cpp — with no per-token cost.

Related models

Reka EdgeRekaai