AI Desk · Model Identity · Compare
Magnum v4 72B · Qwen: Qwen3 Max Thinking
Identity similarity: very highExact tokenizer agreement · 50/50Exact agreement across every measured string is strong evidence of shared tokenizer lineage. It shows shared ancestry, and does not on its own establish that these are the same release.
Magnum v4 72B recordQwen: Qwen3 Max Thinking recordOpen in the interactive bench
Tokenizer fingerprint
50 of 50 mutually clean strings agree| Test string | Magnum v4 72B | Qwen: Qwen3 Max Thinking | Verdict |
|---|---|---|---|
| en-prose | 14 | 14 | match |
| en-long | 16 | 16 | match |
| spaces-20 | 3 | 3 | match |
| spaces-60 | 3 | 3 | match |
| tabs-20 | 3 | 3 | match |
| newlines-20 | 4 | 4 | match |
| mixed-ws | 5 | 5 | 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 | 10 | 10 | match |
| zh-long | 12 | 12 | match |
| zh-rare | 17 | 17 | match |
| ja-kana | 8 | 8 | match |
| ja-kanji | 13 | 13 | match |
| ko | 17 | 17 | match |
| ru | 16 | 16 | match |
| ar | 8 | 8 | match |
| he | 9 | 9 | match |
| hi | 29 | 29 | match |
| th | 12 | 12 | match |
| el | 28 | 28 | match |
| emoji-basic | 5 | 5 | match |
| emoji-skin | 10 | 10 | match |
| emoji-zwj-family | 10 | 10 | match |
| emoji-zwj-x3 | 30 | 30 | match |
| emoji-flags | 8 | 8 | match |
| emoji-prof | 14 | 14 | match |
| math | 26 | 26 | match |
| boxdraw | 17 | 17 | match |
| combining | 5 | 5 | match |
| cjk-ext-b | 12 | 12 | match |
| surrogates | 22 | 22 | match |
| zalgo | 36 | 36 | match |
| rtl-mix | 7 | 7 | match |
| py-code | 25 | 25 | match |
| py-indent | 15 | 15 | match |
| json | 25 | 25 | match |
| html | 16 | 16 | match |
| regex | 44 | 44 | match |
| camel | 5 | 5 | match |
| snake | 6 | 6 | match |
| rare-word-x5 | 31 | 31 | match |
| repeat-tok | 22 | 22 | match |
| base64 | 36 | 36 | match |
| hex | 17 | 17 | match |
| url | 19 | 19 | 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 | Magnum v4 72B | Qwen: Qwen3 Max Thinking | Verdict |
|---|---|---|---|
| Context length | 32768 | 262144 | × differs |
| Max output tokens | 4096 | 65536 | × differs |
| Supported parameters | frequency_penalty, logit_bias, logprobs, max_tokens, min_p, presence_penalty, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, top_a, top_k, top_logprobs, top_p | frequency_penalty, include_reasoning, logprobs, max_tokens, presence_penalty, reasoning, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, top_p | × differs |
| Defaults | {} | {"temperature":null,"top_p":null,"frequency_penalty":null} | × differs |
| Reasoning contract | – | {"mandatory":false} | × differs |
| Modalities | text->text | text->text | ✓ match |
Method: tokenizer fingerprinting, v1, 50 strings. This page is static and dated; the interactive bench compares any two of the 486 catalogue models.