Hanzo Dev
commited on
Commit
·
23f0fa8
1
Parent(s):
77e3bd2
Add YAML metadata and update architecture details (DeepseekV3 MoE 671B/14B active)
Browse files
README.md
CHANGED
|
@@ -1,11 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
# Zen Max - Kimi K2 Thinking Architecture
|
| 2 |
|
| 3 |
**Organization**: [Zen LM](https://zenlm.org) (Hanzo AI × Zoo Labs Foundation)
|
| 4 |
-
**Base Model**: Moonshot AI Kimi K2 Thinking
|
| 5 |
-
**Parameters**:
|
| 6 |
**License**: Apache 2.0
|
| 7 |
**Context Window**: 256K tokens
|
| 8 |
**Thinking Capacity**: 96K-128K thinking tokens per step
|
|
|
|
| 9 |
|
| 10 |
## Model Overview
|
| 11 |
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
tags:
|
| 5 |
+
- deepseek
|
| 6 |
+
- kimi_k2
|
| 7 |
+
- text-generation
|
| 8 |
+
- reasoning
|
| 9 |
+
- agentic
|
| 10 |
+
- tool-calling
|
| 11 |
+
- compressed-tensors
|
| 12 |
+
pipeline_tag: text-generation
|
| 13 |
+
base_model: moonshotai/Kimi-K2-Thinking
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
# Zen Max - Kimi K2 Thinking Architecture
|
| 17 |
|
| 18 |
**Organization**: [Zen LM](https://zenlm.org) (Hanzo AI × Zoo Labs Foundation)
|
| 19 |
+
**Base Model**: Moonshot AI Kimi K2 Thinking (DeepseekV3ForCausalLM)
|
| 20 |
+
**Parameters**: 671B total (384 experts × ~1.75B each, 8 active per token = ~14B)
|
| 21 |
**License**: Apache 2.0
|
| 22 |
**Context Window**: 256K tokens
|
| 23 |
**Thinking Capacity**: 96K-128K thinking tokens per step
|
| 24 |
+
**Architecture**: DeepseekV3 MoE (Mixture of Experts)
|
| 25 |
|
| 26 |
## Model Overview
|
| 27 |
|