Reka Flash 3

Reka Flash 3 places in the top three on HumanEval.

Reka Released March 10th, 2025 Compare with others

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Reka Mar 2025

Reka Flash 3

Open weights

129.0 #282 of 404

Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.

LiveCodeBench v5 62% #24 /47
BALROG 29.2% #16 /33
MMLU-Pro 65% #137 /167

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

Benchmark	Score	Rank	Source
Stick to Your Role!	0.39	#10 of 15	Official leaderboard
Creative Writing v3	54.8	#97 of 120	Official leaderboard
MMLU-Pro	65%	#137 of 167	Lab model card
Longform Creative Writing (EQ-Bench)	41.2	#98 of 119	Official leaderboard

Math and science

Benchmark	Score	Rank	Source
GPQA Diamond	61.1%	#218 of 302	Lab model card
MATH-500	90.6%	#51 of 70	Lab model card
AIME 2024	70%	#65 of 88	Lab model card
AIME 2025	50%	#131 of 143	Lab model card

Coding

Benchmark	Score		Rank	Source
HumanEval	96.3%		#3 of 101	Lab model card
LiveCodeBench v5	62%		#24 of 47	Lab model card

Agents and tools

Benchmark	Score		Rank	Source
BALROG	29.2%		#16 of 33	Official leaderboard
SnakeBench	11		#182 of 188	Official leaderboard

Its lab's numbers put it at 129.0, number 282 of 404

The Self-Reported Index fits one score to the 7 capability results Reka reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 126.7 to 131.4. How we compute it.

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

Reasoning with Reka Flash 3 (Reka news) Mar 10, 2025