Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,13 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
---
|
| 4 |
+
# BanBTP Model Family
|
| 5 |
+
|
| 6 |
+
| Model | Base Architecture | Parameters | Context | Status | Notes |
|
| 7 |
+
|-------|-------------------|------------|---------|--------|-------|
|
| 8 |
+
| BanBTP 124M | GPT-2 (from scratch) | 124M | 1024 | Archaic | Baseline, useless |
|
| 9 |
+
| BanBTP 783M | DialoGPT | 783M | 1024 | Archaic | Repeats and breaks |
|
| 10 |
+
| BanBTP V2 | Gemma 4 | 31B | 1M | Obsolete | Heavy — needs 200GB+ RAM. Inconsistent reasoning. |
|
| 11 |
+
| BanBTP V3 | Gemma 4 | 5B | 1M | Current flagship | Stable, remembers facts, thinks through problems |
|
| 12 |
+
| BanBTP V4 (coder) | Gemma 4 | 5B | 1M | Public variant | Fine-tuned for coding + reasoning only |
|
| 13 |
+
| **BanBTP INFINITE** | **Custom architecture (distilled from V3 + V4)** | **Unknown** | **2M** | **Closed source / Flagship** | Combines chat + coding. Beats V2 easily. |
|