AI Desk · Model Identity · Compare
Reka Edge · Reka Flash 3
Identity similarity: lowDivergent fingerprints · 32/46These models tokenise differently. Divergence on the high-information strings (CJK, emoji, digits) separates unrelated vocabularies within a handful of samples.
Reka Edge recordReka Flash 3 recordOpen in the interactive bench
Tokenizer fingerprint
32 of 46 mutually clean strings agree| Test string | Reka Edge | Reka Flash 3 | Verdict |
|---|---|---|---|
| en-prose | 14 | 14 | match |
| en-long | 15 | 15 | match |
| spaces-20 | 3 | 3 | match |
| spaces-60 | 3 | 3 | match |
| tabs-20 | 3 | 4 | differs |
| newlines-20 | 4 | 5 | differs |
| mixed-ws | 5 | 5 | match |
| digits-9 | 4 | 4 | match |
| digits-12 | 5 | 6 | differs |
| digits-30 | 11 | 16 | differs |
| digits-sep | 8 | 8 | match |
| float-long | 10 | 11 | differs |
| zh-common | 19 | 19 | match |
| zh-long | 26 | 26 | match |
| zh-rare | 24 | 24 | match |
| ja-kana | 15 | 15 | match |
| ja-kanji | 17 | 17 | match |
| ko | 19 | 19 | match |
| ru | 16 | 16 | match |
| ar | 25 | 25 | match |
| he | 24 | 24 | match |
| hi | 32 | 35 | differs |
| th | 26 | 27 | differs |
| el | 28 | 28 | match |
| emoji-basic | 9 | 9 | match |
| emoji-skin | 29 | 29 | match |
| emoji-zwj-family | 17 | 17 | match |
| emoji-zwj-x3 | 53 | 53 | match |
| emoji-flags | 24 | 24 | match |
| emoji-prof | 28 | 28 | match |
| math | 31 | 31 | match |
| boxdraw | 28 | x | – |
| combining | 14 | 14 | match |
| cjk-ext-b | 14 | x | – |
| surrogates | 40 | 39 | differs |
| zalgo | 36 | x | – |
| rtl-mix | 11 | 11 | match |
| py-code | 25 | 28 | differs |
| py-indent | 15 | 18 | differs |
| json | 20 | 23 | differs |
| html | 17 | 17 | match |
| regex | 44 | 47 | differs |
| camel | 5 | 5 | match |
| snake | 6 | 11 | differs |
| rare-word-x5 | 31 | 31 | match |
| repeat-tok | 21 | 21 | match |
| base64 | 35 | 35 | match |
| hex | 11 | x | – |
| url | 17 | 21 | differs |
| uuid | 28 | 28 | match |
Values are the marginal prompt-token cost of each string with the model's fixed overhead subtracted. An x marks a row excluded as corrupt for that model.
API surface
The declared serving contracts differ| Field | Reka Edge | Reka Flash 3 | Verdict |
|---|---|---|---|
| Context length | 16384 | 65536 | × differs |
| Max output tokens | 14745 | 58982 | × differs |
| Supported parameters | frequency_penalty, logprobs, max_tokens, presence_penalty, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, top_p | frequency_penalty, include_reasoning, logprobs, max_tokens, presence_penalty, reasoning, seed, stop, structured_outputs, temperature, top_k, top_logprobs, top_p | × differs |
| Defaults | {"temperature":null,"top_p":null,"top_k":null,"frequency_penalty":null,"presence_penalty":null,"repetition_penalty":null} | {"temperature":null,"top_p":null,"top_k":null,"frequency_penalty":null,"presence_penalty":null,"repetition_penalty":null} | ✓ match |
| Reasoning contract | {"mandatory":false} | {"mandatory":true} | × differs |
| Modalities | text+image+video->text | text->text | × differs |
Method: tokenizer fingerprinting, v1, 50 strings. This page is static and dated; the interactive bench compares any two of the 453 catalogue models.