Skip to content
llmwaves

Reka Flash 3

Reka AI · released Mar 12, 2025

Open weightsReasoningStructured output

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...

Specification

Context window
66K
Max output
66K
Knowledge cutoff
Jan 31, 2025
Parameters
20.9B
Licence
apache-2.0
Serving providers
1
Moderated
No

Intelligence

3.7

12th percentile

Coding

Coding Index

Agentic

Agentic Index

Output speed

90 t/s

Median across providers

Latency

1.21s

Time to first token

Cost per task

$0.02

Estimated

Benchmarks

Where the score comes from

The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.

Evaluation scores

Percentage correct · higher is better

  • MMLU-Pro
    66.9%
  • GPQA Diamond
    52.9%
  • LiveCodeBench
    43.5%
  • AIME 2025
    33.7%
  • IFBench
    30.4%
  • SciCode
    26.7%
  • Humanity's Last Exam
    4.4%
  • τ²-bench (Telecom)
    0.0%
  • Terminal-Bench Hard
    0.0%
  • AA-LCR (long context)
    0.0%

An evaluation missing from this list was not run for this model — it is not a zero.

View as table
EvaluationScore
MMLU-Pro66.9%
GPQA Diamond52.9%
LiveCodeBench43.5%
AIME 202533.7%
IFBench30.4%
SciCode26.7%
Humanity's Last Exam4.4%
τ²-bench (Telecom)0.0%
Terminal-Bench Hard0.0%
AA-LCR (long context)0.0%

Against its peers

Intelligence Index · this model highlighted, nearest peers in grey

Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.

View as table
ModelIntelligence
Olmo 3 32B Think6.1
Hermes 3 70B Instruct4.8
Phi 44.6
Mistral Large4.1
Mixtral 8x22B Instruct4.0
Reka Flash 33.7
Claude 3 Haiku3.5
GPT-3.5 Turbo3.2

Percentile among all indexed models

Intelligence12th
Mathematics34th
Terminal work0th

Pricing

What it costs to run

List prices per million tokens, plus what one representative task works out to.

List price

Input / 1M tokens
$0.1
Output / 1M tokens
$0.2
Cached input / 1M
Not offered
Blended 3:1
$0.125

One task, estimated

$0.02

Input tokens
50,000
Output tokens
80,000
Profile
Reasoning

Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).

Serving providers

Speed and latency figures are medians across these providers, so a widely-served model reports a blend rather than any single endpoint.