mradermacher commited on
Commit
7536a57
·
verified ·
1 Parent(s): baa857a

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +7 -0
README.md CHANGED
@@ -35,6 +35,7 @@ more details, including on how to concatenate multi-part files.
35
 
36
  | Link | Type | Size/GB | Notes |
37
  |:-----|:-----|--------:|:------|
 
38
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ1_M.gguf) | i1-IQ1_M | 1.9 | mostly desperate |
39
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 2.1 | |
40
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ2_XS.gguf) | i1-IQ2_XS | 2.3 | |
@@ -42,15 +43,21 @@ more details, including on how to concatenate multi-part files.
42
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ2_M.gguf) | i1-IQ2_M | 2.6 | |
43
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q2_K.gguf) | i1-Q2_K | 2.8 | IQ3_XXS probably better |
44
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 2.9 | lower quality |
 
45
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q3_K_S.gguf) | i1-Q3_K_S | 3.3 | IQ3_XS probably better |
 
46
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ3_M.gguf) | i1-IQ3_M | 3.4 | |
47
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q3_K_M.gguf) | i1-Q3_K_M | 3.6 | IQ3_S probably better |
48
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q3_K_L.gguf) | i1-Q3_K_L | 3.9 | IQ3_M probably better |
49
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ4_XS.gguf) | i1-IQ4_XS | 4.0 | |
50
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q4_0_4_4.gguf) | i1-Q4_0_4_4 | 4.2 | fast on arm, low quality |
 
 
 
51
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q4_K_S.gguf) | i1-Q4_K_S | 4.2 | optimal size/speed/quality |
52
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q4_K_M.gguf) | i1-Q4_K_M | 4.5 | fast, recommended |
53
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q5_K_S.gguf) | i1-Q5_K_S | 5.1 | |
 
54
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q6_K.gguf) | i1-Q6_K | 6.0 | practically like static Q6_K |
55
 
56
  Here is a handy graph by ikawrakow comparing some lower-quality quant
 
35
 
36
  | Link | Type | Size/GB | Notes |
37
  |:-----|:-----|--------:|:------|
38
+ | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ1_S.gguf) | i1-IQ1_S | 1.7 | for the desperate |
39
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ1_M.gguf) | i1-IQ1_M | 1.9 | mostly desperate |
40
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 2.1 | |
41
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ2_XS.gguf) | i1-IQ2_XS | 2.3 | |
 
43
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ2_M.gguf) | i1-IQ2_M | 2.6 | |
44
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q2_K.gguf) | i1-Q2_K | 2.8 | IQ3_XXS probably better |
45
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 2.9 | lower quality |
46
+ | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ3_XS.gguf) | i1-IQ3_XS | 3.1 | |
47
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q3_K_S.gguf) | i1-Q3_K_S | 3.3 | IQ3_XS probably better |
48
+ | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ3_S.gguf) | i1-IQ3_S | 3.3 | beats Q3_K* |
49
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ3_M.gguf) | i1-IQ3_M | 3.4 | |
50
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q3_K_M.gguf) | i1-Q3_K_M | 3.6 | IQ3_S probably better |
51
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q3_K_L.gguf) | i1-Q3_K_L | 3.9 | IQ3_M probably better |
52
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-IQ4_XS.gguf) | i1-IQ4_XS | 4.0 | |
53
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q4_0_4_4.gguf) | i1-Q4_0_4_4 | 4.2 | fast on arm, low quality |
54
+ | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q4_0_4_8.gguf) | i1-Q4_0_4_8 | 4.2 | fast on arm+i8mm, low quality |
55
+ | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q4_0_8_8.gguf) | i1-Q4_0_8_8 | 4.2 | fast on arm+sve, low quality |
56
+ | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q4_0.gguf) | i1-Q4_0 | 4.2 | fast, low quality |
57
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q4_K_S.gguf) | i1-Q4_K_S | 4.2 | optimal size/speed/quality |
58
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q4_K_M.gguf) | i1-Q4_K_M | 4.5 | fast, recommended |
59
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q5_K_S.gguf) | i1-Q5_K_S | 5.1 | |
60
+ | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q5_K_M.gguf) | i1-Q5_K_M | 5.2 | |
61
  | [GGUF](https://huggingface.co/mradermacher/MBX-7B-v3-i1-GGUF/resolve/main/MBX-7B-v3.i1-Q6_K.gguf) | i1-Q6_K | 6.0 | practically like static Q6_K |
62
 
63
  Here is a handy graph by ikawrakow comparing some lower-quality quant