Reka: Reka Flash 3

reka-flash-3

Reka Flash 3 is a 21B general-purpose reasoning LLM which excels at chat, coding, and instruction-following while maintaining a level of performance competitive with significantly larger models. Note: Reka Flash 3 is primarily an English model with limited multilingual understanding capabilities.

Modalities: →
In / out price: $0.10 / $0.20 per 1M
Context: 64k
Max output: 65,536 tokens

ChatStreamingReasoningStructured outputs

Uptime

Uptime is the share of the past 24 hours in which the model was serving requests. Availability is the share of requests that succeeded, ignoring caller-side errors such as validation or rate limits.

Uptime (24h)i

100.00%

Hours with traffic in which requests were served

Availability (24h)i

100.00%

2,012 requests counted

Availability over the last 24 hours

Availability 100.00%

Now

Hourly availability

Performance

Throughput is how fast the model writes (tokens per second, higher is better). Latency is time to first token: how long before you see anything appear (lower is better). Both are measured on streamed requests over the last 24 hours.

Throughputi

116tok/s

Average · 2,012 streamed requests

Latencyi

0.21 s

P50 time to first token · avg 0.29 s

Throughput

Latency

Activity

Tokens processed by this model over the last 24 hours, split into prompt, answer and reasoning tokens. Reasoning tokens are billed as output.

Requests (24h)

2,012

Successful chat completions

Tokens (24h)

14.2M

3.8M prompt · 10.4M answer · 0 reasoning

Avg tokens / request

7,066

Prompt plus completion

Tokens per hour

PromptAnswerReasoning

This model does not report reasoning tokens separately; its thinking is counted in the answer slice.

Loading live metrics…

Reka: Reka Flash 3

Uptime

Availability over the last 24 hours

Hourly availability

Performance

Throughput

Latency

Activity

Tokens per hour

Next steps

Uptime

Availability over the last 24 hours

Hourly availability

Performance

Throughput

Latency

Activity

Tokens per hour