bharath-inferock bharathkoneti commited on
Commit
ecac1e5
·
1 Parent(s): f631ea7

Add the Dane model card (#1)

Browse files

- Add the Dane model card (3db7678d8b5716a44ff1cd80b1c5b350392a1621)


Co-authored-by: Bharath Koneti <bharathkoneti@users.noreply.huggingface.co>

Files changed (1) hide show
  1. README.md +48 -0
README.md ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: other
3
+ license_name: inferock-terms
4
+ pipeline_tag: text-generation
5
+ inference: false
6
+ language:
7
+ - en
8
+ tags:
9
+ - hybrid-model
10
+ - frontier
11
+ - inferock
12
+ - dane
13
+ ---
14
+
15
+ # DANE: Decision Aware Neural Engine
16
+
17
+ Dane is a hybrid, frontier class model. It beats the frontier and delivers the most reliable answer on every request.
18
+
19
+ ## What Dane is
20
+
21
+ Dane is a hybrid model made by Inferock.
22
+ Dane scores 79.4 on HLE Diamond, 15 percent above the Artificial Analysis board leader.
23
+ Dane is API-only: you reach it through the Inferock API as the model id `dane`, and no weights are published on the Hub.
24
+
25
+ ## Results
26
+
27
+ **HLE Diamond: Dane 79.4, index leader 69.1**
28
+
29
+ HLE Diamond is the 1,000-question Diamond set of Humanity's Last Exam (dataset cais/hle-diamond), run with tools and graded as the Artificial Analysis board grades: the official HLE judge prompt, GPT-5.6 Luna (medium) as the judge, pass@1. Score is right over judged: 794 of 1,000; a question with no answer counts as wrong. Run by Inferock, October 30 to November 4, 2026.
30
+
31
+ ## Use Dane
32
+
33
+ Dane answers a standard chat completions request with the model id `dane`.
34
+ Sign up at https://inferock.ai, activate Serve, create a key, and use the base URL from your account.
35
+ Pricing: 8 USD per million input tokens, 30 USD per million output tokens (billed tokens).
36
+ Context window and modalities: as listed on https://inferock.ai/models/.
37
+
38
+ ## FAQ
39
+
40
+ How does Dane do it? Dane is built on one rule: a model's answer is a prediction, not a verdict. Everything else is how we measure it, and the results are on the releases page.
41
+
42
+ ## Links
43
+
44
+ - https://inferock.ai
45
+ - https://inferock.ai/dane/
46
+ - https://inferock.ai/docs/dane/
47
+ - https://inferock.ai/docs/pricing/
48
+ - Contact: see https://inferock.ai