File size: 2,617 Bytes
862f28a
 
 
 
 
 
 
 
 
 
 
 
 
cfd3732
862f28a
 
 
cfd3732
862f28a
cfd3732
862f28a
cfd3732
862f28a
cfd3732
862f28a
cfd3732
862f28a
cfd3732
862f28a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
cfd3732
e633844
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
---
license: cc-by-4.0
tags:
  - iceman-contributor
  - stealth-model
  - stealthmark
  - coding
  - reasoning
  - long-context
  - tool-use
thumbnail: https://stealthmodels.com/iceman-contributor/assets/social-card.png
---

<img src="https://huggingface.co/stealth-model/iceman-contributor/resolve/main/panels/hero.webp" alt="A boy made of cracked ice lunging beside the name Iceman Contributor, 417 billion tokens on day one, two benchmark results and a 23 of 23 tokenizer match.">

**Iceman Contributor counts tokens exactly like MAI Thinking 1 and Glyph Cluster, on 23 of 23 phrases, and no company has claimed it yet.**

By [Stealth Models](https://stealthmodels.com/).

<img src="https://huggingface.co/stealth-model/iceman-contributor/resolve/main/panels/verdict.webp" alt="StealthMark 1.8.10 results for Iceman Contributor: benchmark score 19.2, General Intelligence 11.7, Visual Intelligence 31.0. Strongest at composing a whole scene: 46.4 on Beach.">

<img src="https://huggingface.co/stealth-model/iceman-contributor/resolve/main/panels/results.webp" alt="Every StealthMark result of Iceman Contributor: Beach 46.4, Selfie 39.3, Raccoon 29.8, Ground Truth 20.2, Mona Lisa 19.3, Newton's Mirror 10.1, Latch 3.6, Foresight 3.2.">

[Compare it with Glyph Cluster](https://stealthmodels.com/compare/glyph-cluster-vs-iceman-contributor)

<img src="https://huggingface.co/stealth-model/iceman-contributor/resolve/main/panels/case.webp" alt="The case file. First seen 24 Oct 2026, revealed not yet, developer unknown, context 262,144 tokens, input Text, tool calls Yes, speed in our tests about 46 tok/s. 24 Oct 2026: Appears without an announcement: about 417 billion tokens on its first full day. 24 Oct 2026: A post on X matches its token counts to Microsoft’s MAI Thinking 1. 25 Oct 2026, our finding: Our test: the same tokenizer family as Glyph Cluster, our MAI lead.">

[See the token-count test](https://stealthmodels.com/iceman-contributor#tokenizer)

<details><summary>Facts as text</summary>

| Fact | Value |
|---|---|
| First seen | 24 Oct 2026 |
| Revealed | not yet |
| Developer | unknown |
| Context | 262,144 tokens |
| Input | Text |
| Tool calls | Yes |
| Speed in our tests | about 46 tok/s |
| Access | Active; prompts may be used for training |

</details>

## Watch
[![Iceman Contributor on the same benchmark tasks, and the case for who made it](https://img.youtube.com/vi/l5hdM8lsPfI/hqdefault.jpg)](https://www.youtube.com/watch?v=l5hdM8lsPfI)

---
[StealthMark leaderboard](https://huggingface.co/spaces/stealth-model/stealthmark)<br>
Text and images: CC BY 4.0, stealthmodels.com.