Reka Flash 3

rekaai/reka-flash-3 · RekaaiNot yet established

Declared family: Unknown
Entered catalogue: 12 March 2025
Last tested: 21 August 2026
Evidence confidence: Unknown

Compare with Reka Edge

Report a correction

Identity statement

Reka Flash 3 has a clean measured fingerprint that currently sits alone: no other measured model shares its signature.

What we checked

Four different kinds of evidence, and what each one showed

We checked	What we found	Where it came from
How it counts tokens	Counted, using the 46 strings that came back reliably	we measured it
How its API is set up	A combination of settings no other listing in the catalogue uses	api_5a786c3d
How much it can read and write	Reads up to 65,536 tokens at onceOpenRouter's listing shows 58,982 output tokens, which is 90% of the context window, the figure it carries where the provider declares no maximum.	its own listing
Whether it thinks before answering	Always reasons, and cannot be turned off	its own listing
Who served our requests	Reka	we measured it, 21 August 2026

Closest measured models

Agreement across the mutually clean test strings

Reka Edgerekaai/reka-edge32/46 · 70%IBM: Granite 4.2 8Bibm-granite/granite-4.2-8b26/46 · 57%Inference.net: Schematron V2 Turboinference-net/schematron-v2-turbo26/46 · 57%OpenAI: GPT-3.5 Turbo Instructopenai/gpt-3.5-turbo-instruct14/38 · 37%IBM: Granite 4.0 Microibm-granite/granite-4.0-h-micro15/46 · 33%IBM: Granite 4.1 8Bibm-granite/granite-4.1-8b15/46 · 33%

Click a row to open the full comparison.

The measured fingerprint

What each test string cost this model, in tokens

We sent Reka Flash 3 fifty short pieces of text and recorded what each one cost it in tokens. The bars below are those costs. Two models built on the same tokenizer produce the same bars; a model built on a different one produces a different set, which is what makes this a fingerprint.

All 50 rows, 4 excluded as corrupt

English and whitespace

en-prose14

en-long15

spaces-203

spaces-603

tabs-204

newlines-205

mixed-ws5

Digits

digits-94

digits-126

digits-3016

digits-sep8

float-long11

CJK

zh-common19

zh-long26

zh-rare24

ja-kana15

ja-kanji17

ko19

Other scripts

ru16

ar25

he24

hi35

th27

el28

Emoji

emoji-basic9

emoji-skin29

emoji-zwj-family17

emoji-zwj-x353

emoji-flags24

emoji-prof28

Rare Unicode

math31

boxdrawx

combining14

cjk-ext-bx

surrogates39

zalgox

rtl-mix11

Code

py-code28

py-indent18

json23

html17

regex47

camel5

snake11

Repetition and encodings

rare-word-x531

repeat-tok21

base6435

hexx

url21

uuid28

An x marks a row excluded as corrupt (caching or provider interference detected during measurement). Overhead subtracted: 12 prompt tokens.

Declared record

What the catalogue claims about this model

Reka Flash 3 is a general-purpose, instruction-tuned large language model with 21 billion parameters, developed by Reka. It excels at general chat, coding tasks, instruction-following, and function calling. Featuring a...

Modalities: text->text
Declared tokenizer: Other
Prompt price: $0.10 / M tokens
Completion price: $0.20 / M tokens

Supported parameters

frequency_penaltyinclude_reasoninglogprobsmax_tokenspresence_penaltyreasoningseedstopstructured_outputstemperaturetop_ktop_logprobstop_p

defaults: {"temperature":null,"top_p":null,"top_k":null,"frequency_penalty":null,"presence_penalty":null,"repetition_penalty":null}

Catalogue entry Weights on Hugging Face

History

Every observation, kept as taken

12 March 2025Enters the OpenRouter catalogue with no declared family.
21 August 2026Fingerprinted in the YFarmX catalogue sweep · 46 of 50 strings measured clean.