AI Desk · Model Identity · Compare
Baidu: ERNIE 4.5 VL 424B A47B · Mancer: Weaver (alpha)
Identity similarity: lowDivergent fingerprints · 24/50These models tokenise differently. Divergence on the high-information strings (CJK, emoji, digits) separates unrelated vocabularies within a handful of samples.
Baidu: ERNIE 4.5 VL 424B A47B recordMancer: Weaver (alpha) recordOpen in the interactive bench
Tokenizer fingerprint
24 of 50 mutually clean strings agree| Test string | Baidu: ERNIE 4.5 VL 424B A47B | Mancer: Weaver (alpha) | Verdict |
|---|---|---|---|
| en-prose | 14 | 16 | differs |
| en-long | 15 | 18 | differs |
| spaces-20 | 4 | 4 | match |
| spaces-60 | 6 | 6 | match |
| tabs-20 | 22 | 22 | match |
| newlines-20 | 22 | 22 | match |
| mixed-ws | 9 | 9 | match |
| digits-9 | 9 | 9 | match |
| digits-12 | 12 | 12 | match |
| digits-30 | 30 | 30 | match |
| digits-sep | 12 | 12 | match |
| float-long | 22 | 22 | match |
| zh-common | 8 | 22 | differs |
| zh-long | 10 | 34 | differs |
| zh-rare | 27 | 27 | match |
| ja-kana | 10 | 16 | differs |
| ja-kanji | 13 | 16 | differs |
| ko | 12 | 29 | differs |
| ru | 13 | 14 | differs |
| ar | 12 | 29 | differs |
| he | 12 | 26 | differs |
| hi | 14 | 31 | differs |
| th | 14 | 26 | differs |
| el | 13 | 33 | differs |
| emoji-basic | 20 | 20 | match |
| emoji-skin | 22 | 40 | differs |
| emoji-zwj-family | 19 | 19 | match |
| emoji-zwj-x3 | 57 | 57 | match |
| emoji-flags | 8 | 32 | differs |
| emoji-prof | 35 | 35 | match |
| math | 44 | 26 | differs |
| boxdraw | 41 | 17 | differs |
| combining | 10 | 10 | match |
| cjk-ext-b | 16 | 16 | match |
| surrogates | 54 | 54 | match |
| zalgo | 36 | 35 | differs |
| rtl-mix | 6 | 13 | differs |
| py-code | 28 | 28 | match |
| py-indent | 19 | 19 | match |
| json | 28 | 28 | match |
| html | 17 | 17 | match |
| regex | 44 | 52 | differs |
| camel | 6 | 8 | differs |
| snake | 11 | 12 | differs |
| rare-word-x5 | 26 | 45 | differs |
| repeat-tok | 21 | 40 | differs |
| base64 | 31 | 41 | differs |
| hex | 15 | 20 | differs |
| url | 23 | 23 | match |
| uuid | 36 | 36 | 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 | Baidu: ERNIE 4.5 VL 424B A47B | Mancer: Weaver (alpha) | Verdict |
|---|---|---|---|
| Context length | 123000 | 8000 | × differs |
| Max output tokens | 16000 | 6000 | × differs |
| Supported parameters | frequency_penalty, include_reasoning, max_tokens, presence_penalty, reasoning, repetition_penalty, seed, stop, temperature, top_k, top_p | frequency_penalty, logit_bias, logprobs, max_tokens, min_p, presence_penalty, repetition_penalty, response_format, seed, stop, temperature, top_a, top_k, top_logprobs, top_p | × differs |
| Defaults | {} | {} | ✓ match |
| Reasoning contract | {"mandatory":false} | – | × differs |
| Modalities | text+image->text | text->text | × differs |
Method: tokenizer fingerprinting, v1, 50 strings. This page is static and dated; the interactive bench compares any two of the 486 catalogue models.