U

Unknown: reka-flash-3

reka-flash-3
65.5K context59.0K outReasoningStructured
Released Mar 12, 2025Knowledge cutoff Jan 2025Updated Nov 6, 2026

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 32K context length and optimized through reinforcement learning (RLOO), it provides competitive performance comparable to proprietary models within a smaller parameter footprint. Ideal for low-latency, local, or on-device deployments, Reka Flash 3 is compact, supports efficient quantization (down to 11GB at 4-bit precision), and employs explicit reasoning tags ("") to indicate its internal thought process. Reka Flash 3 is primarily an English model with limited multilingual understanding capabilities. The model weights are released under the Apache 2.0 license.

Mode chatTokenizer OtherRekaAI/reka-flash-3

Pricing

Input price
$0.11/ 1M tokens
Output price
$0.22/ 1M tokens
Context window 65.5K tokensCompatible endpoints openaiVendor Unknown

Uptime

Performance

Loading performance data...

Usage & Ranking

Loading usage...

Supported parameters

All providers = every upstream serving this model supports it. Some providers = depends on which upstream handles the request. Default = the value sent when you leave the parameter unset.

ParameterProvidersDefault
frequency_penaltyAll providersNot sent by default
include_reasoningAll providers-
logprobsAll providers-
max_tokensAll providers-
presence_penaltyAll providersNot sent by default
reasoningAll providers-
repetition_penaltySome providersNot sent by default
seedAll providers-
stopAll providers-
structured_outputsAll providers-
temperatureAll providersNot sent by default
top_kAll providersNot sent by default
top_logprobsAll providers-
top_pAll providersNot sent by default

Frequently asked questions

Similar models