AI Desk · AI Fingerprinting · Compare
Anthropic: Claude Fable 5.1 · IBM: Granite 4.0 Micro
Identity similarity: lowDivergent fingerprints · 15/49These models tokenise differently. The rows that part here are the English prose, whitespace, CJK, non-Latin script and emoji strings. Divergence on high-information strings separates unrelated vocabularies within a handful of samples.
Anthropic: Claude Fable 5.1 recordIBM: Granite 4.0 Micro recordOpen in the interactive bench
Tokenizer fingerprint
15 of 49 mutually clean strings agree| Test string | Anthropic: Claude Fable 5.1 | IBM: Granite 4.0 Micro | Verdict |
|---|---|---|---|
| en-prose | 24 | 14 | differs |
| en-long | 28 | 16 | differs |
| spaces-20 | 4 | 3 | differs |
| spaces-60 | 6 | 3 | differs |
| tabs-20 | 4 | 3 | differs |
| newlines-20 | 4 | 4 | match |
| mixed-ws | 9 | 5 | differs |
| digits-9 | 3 | 3 | match |
| digits-12 | 4 | 4 | match |
| digits-30 | 10 | 10 | match |
| digits-sep | 7 | 7 | match |
| float-long | 9 | 9 | match |
| zh-common | 18 | 18 | match |
| zh-long | 24 | 25 | differs |
| zh-rare | 20 | 23 | differs |
| ja-kana | 12 | 14 | differs |
| ja-kanji | 16 | 17 | differs |
| ko | 19 | 19 | match |
| ru | 14 | 16 | differs |
| ar | 21 | 25 | differs |
| he | 18 | 24 | differs |
| hi | 23 | 32 | differs |
| th | 25 | 25 | match |
| el | 22 | 29 | differs |
| emoji-basic | 10 | 10 | match |
| emoji-skin | 25 | 30 | differs |
| emoji-zwj-family | 14 | 18 | differs |
| emoji-zwj-x3 | 42 | 54 | differs |
| emoji-flags | 24 | 24 | match |
| emoji-prof | 27 | 29 | differs |
| math | 46 | 32 | differs |
| boxdraw | 28 | 28 | match |
| combining | 6 | 14 | differs |
| cjk-ext-b | 16 | 13 | differs |
| surrogates | 54 | 40 | differs |
| zalgo | 36 | 36 | match |
| rtl-mix | 11 | 11 | match |
| py-code | 30 | 25 | differs |
| py-indent | 19 | 15 | differs |
| json | 29 | 20 | differs |
| html | 18 | 16 | differs |
| regex | 63 | 44 | differs |
| camel | 16 | 5 | differs |
| snake | 14 | 6 | differs |
| rare-word-x5 | 55 | 31 | differs |
| repeat-tok | 60 | 22 | differs |
| base64 | x | 35 | – |
| hex | 18 | 11 | differs |
| url | 26 | 17 | differs |
| uuid | 27 | 27 | 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 | Anthropic: Claude Fable 5.1 | IBM: Granite 4.0 Micro | Verdict |
|---|---|---|---|
| Context length | 1000000 | 131000 | × differs |
| Max output tokens | 128000 | – | × differs |
| Supported parameters | include_reasoning, max_completion_tokens, max_tokens, reasoning, reasoning_effort, response_format, stop, structured_outputs, tool_choice, tools, verbosity | frequency_penalty, logit_bias, logprobs, max_tokens, min_p, presence_penalty, repetition_penalty, response_format, seed, stop, temperature, top_k, top_logprobs, top_p | × differs |
| Defaults | {} | {"temperature":null,"top_p":null,"frequency_penalty":null} | × differs |
| Reasoning contract | {"mandatory":true,"supported_efforts":["max","xhigh","high","medium","low"],"default_effort":"high"} | – | × differs |
| Modalities | text+image+file->text | text->text | × differs |
IBM: Granite 4.0 Micro: OpenRouter's listing shows 117,900 output tokens, which is 90% of the context window, the figure it carries where the provider declares no maximum.
Method: tokenizer fingerprinting, v1, 50 strings. This page is static and dated; the interactive bench compares any two of the 338 measured models.