### 🔀 [#109](https://github.com/ikawrakow/ik_llama.cpp/pull/109) - Bitnet CUDA improvements

| **Author** | `ikawrakow` |
| :--- | :--- |
| **State** | ❌ **Closed** |
| **Created** | 2024-10-26 |
| **Updated** | 2024-10-26 |

---

#### Description

`IQ1_BN` TG-128 on RTX-4080 goes to 340 t/s up from 318 t/s.
On the front page the performance listed for `IQ1_BN` on CUDA is 301 t/s, so a pretty nice improvement since then.