73 lines
3.7 KiB
Markdown
73 lines
3.7 KiB
Markdown
---
|
||
license: mit
|
||
base_model:
|
||
- WeiboAI/VibeThinker-1.5B
|
||
language:
|
||
- en
|
||
pipeline_tag: text-generation
|
||
library_name: transformers
|
||
tags:
|
||
- gpqa
|
||
- text-generation-inference
|
||
- math
|
||
- code
|
||
---
|
||
|
||
# **VibeThinker-1.5B-f32-GGUF**
|
||
|
||
> [VibeThinker-1.5B](https://huggingface.co/WeiboAI/VibeThinker-1.5B) is a 1.5‑billion‑parameter dense language model from WeiboAI, fine‑tuned from Qwen2.5-Math-1.5B and purpose‑built for competitive math and algorithmic coding problems, where it delivers frontier‑level reasoning despite its small size. Trained with the “Spectrum‑to‑Signal Principle” framework that first maximizes solution diversity in supervised fine‑tuning and then reinforces correct reasoning paths via reinforcement learning, it achieves scores of 80.3, 74.4, and 50.4 on AIME24, AIME25, and HMMT25 respectively—surpassing the much larger DeepSeek R1—and reaches 55.9/51.1 on LiveCodeBench v5/v6, rivaling or beating larger models like Magistral Medium. The model is fully open source under the MIT license, trained with a reported post‑training cost of about $7,800, and is recommended for use specifically on math and coding tasks with long output lengths (up to around 40k tokens) using moderate sampling temperatures.
|
||
|
||
## Model Files
|
||
|
||
### VibeThinker-1.5B [GGUF]
|
||
|
||
| File Name | Quant Type | File Size |
|
||
| - | - | - |
|
||
| VibeThinker-1.5B.BF16.gguf | BF16 | 3.56 GB |
|
||
| VibeThinker-1.5B.F32.gguf | F32 | 7.11 GB |
|
||
| VibeThinker-1.5B.f16.gguf | F16 | 3.56 GB |
|
||
| VibeThinker-1.5B.IQ4_XS.gguf | IQ4_XS | 1.03 GB |
|
||
| VibeThinker-1.5B.Q2_K.gguf | Q2_K | 753 MB |
|
||
| VibeThinker-1.5B.Q3_K_L.gguf | Q3_K_L | 980 MB |
|
||
| VibeThinker-1.5B.Q3_K_M.gguf | Q3_K_M | 924 MB |
|
||
| VibeThinker-1.5B.Q3_K_S.gguf | Q3_K_S | 861 MB |
|
||
| VibeThinker-1.5B.Q4_K_M.gguf | Q4_K_M | 1.12 GB |
|
||
| VibeThinker-1.5B.Q4_K_S.gguf | Q4_K_S | 1.07 GB |
|
||
| VibeThinker-1.5B.Q5_K_M.gguf | Q5_K_M | 1.29 GB |
|
||
| VibeThinker-1.5B.Q5_K_S.gguf | Q5_K_S | 1.26 GB |
|
||
| VibeThinker-1.5B.Q6_K.gguf | Q6_K | 1.46 GB |
|
||
| VibeThinker-1.5B.Q8_0.gguf | Q8_0 | 1.89 GB |
|
||
| VibeThinker-1.5B.i1-IQ1_M.gguf | i1-IQ1_M | 541 MB |
|
||
| VibeThinker-1.5B.i1-IQ1_S.gguf | i1-IQ1_S | 513 MB |
|
||
| VibeThinker-1.5B.i1-IQ2_M.gguf | i1-IQ2_M | 701 MB |
|
||
| VibeThinker-1.5B.i1-IQ2_S.gguf | i1-IQ2_S | 664 MB |
|
||
| VibeThinker-1.5B.i1-IQ2_XS.gguf | i1-IQ2_XS | 627 MB |
|
||
| VibeThinker-1.5B.i1-IQ2_XXS.gguf | i1-IQ2_XXS | 588 MB |
|
||
| VibeThinker-1.5B.i1-IQ3_M.gguf | i1-IQ3_M | 877 MB |
|
||
| VibeThinker-1.5B.i1-IQ3_S.gguf | i1-IQ3_S | 863 MB |
|
||
| VibeThinker-1.5B.i1-IQ3_XS.gguf | i1-IQ3_XS | 832 MB |
|
||
| VibeThinker-1.5B.i1-IQ3_XXS.gguf | i1-IQ3_XXS | 769 MB |
|
||
| VibeThinker-1.5B.i1-IQ4_NL.gguf | i1-IQ4_NL | 1.07 GB |
|
||
| VibeThinker-1.5B.i1-IQ4_XS.gguf | i1-IQ4_XS | 1.02 GB |
|
||
| VibeThinker-1.5B.i1-Q2_K.gguf | i1-Q2_K | 753 MB |
|
||
| VibeThinker-1.5B.i1-Q2_K_S.gguf | i1-Q2_K_S | 717 MB |
|
||
| VibeThinker-1.5B.i1-Q3_K_L.gguf | i1-Q3_K_L | 980 MB |
|
||
| VibeThinker-1.5B.i1-Q3_K_M.gguf | i1-Q3_K_M | 924 MB |
|
||
| VibeThinker-1.5B.i1-Q3_K_S.gguf | i1-Q3_K_S | 861 MB |
|
||
| VibeThinker-1.5B.i1-Q4_0.gguf | i1-Q4_0 | 1.07 GB |
|
||
| VibeThinker-1.5B.i1-Q4_1.gguf | i1-Q4_1 | 1.16 GB |
|
||
| VibeThinker-1.5B.i1-Q4_K_M.gguf | i1-Q4_K_M | 1.12 GB |
|
||
| VibeThinker-1.5B.i1-Q4_K_S.gguf | i1-Q4_K_S | 1.07 GB |
|
||
| VibeThinker-1.5B.i1-Q5_K_M.gguf | i1-Q5_K_M | 1.29 GB |
|
||
| VibeThinker-1.5B.i1-Q5_K_S.gguf | i1-Q5_K_S | 1.26 GB |
|
||
| VibeThinker-1.5B.i1-Q6_K.gguf | i1-Q6_K | 1.46 GB |
|
||
| VibeThinker-1.5B.imatrix.gguf | imatrix | 2.07 MB |
|
||
|
||
## Quants Usage
|
||
|
||
(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
|
||
|
||
Here is a handy graph by ikawrakow comparing some lower-quality quant
|
||
types (lower is better):
|
||
|
||
 |