Upload README.md with huggingface_hub

README.md

@@ -104,15 +104,14 @@ Numbers below are total memory used (model + KV cache + 0.9 GB recurrent state).
 
 | RAM | Quant | KV cache | Max context | Total used | Vision |
 |---|---|---|---|---:|---|
-| 16 GB | `IQ2_M` | `q4_0` | 32K | 11.4 GB | ✗ |
-| 24 GB | **`IQ3_M`** | `q4_0` | **128K** | **15.0 GB** | ✓ |
+| 16 GB | **`IQ2_M`** | `q4_0` | **32K** | **11.1 GB** | ✗ |
+| 24 GB | **`IQ3_M`** | `q4_0` | **128K** | **16.0 GB** | ✓ |
 | 24 GB | `IQ3_M` | `q4_0` | 180K | 15.9 GB | ✗ |
-| 32 GB | **`Q5_K_M`** | `q4_0` | **262K** | **23.2 GB** | ✗ |
-| 32 GB | `Q4_K_M` | `q4_0` | 262K | 21.2 GB | ✓ |
-| 32 GB | `Q5_K_M` | `q8_0` | 128K | 23.0 GB | ✓ |
-| 48 GB | **`Q6_K`** | `q8_0` | **262K** | **30.4 GB** | ✓ |
-| 48 GB | `Q8_0` | `q8_0` | 262K | 36.4 GB | ✓ |
-| 64+ GB | **`Q8_0`** | `q8_0` | **262K** | **36.4 GB** | ✓ |
+| 32 GB | **`Q5_K_M`** | `q4_0` | **262K** | **23.5 GB** | ✗ |
+| 32 GB | `Q4_K_M` | `q4_0` | 262K | 21.8 GB | ✓ |
+| 32 GB | `Q5_K_M` | `q8_0` | 128K | 23.4 GB | ✗ |
+| 48 GB | **`Q6_K`** | `q8_0` | **262K** | **31.2 GB** | ✓ |
+| 48 GB | `Q8_0` | `q8_0` | 262K | 37.3 GB | ✓ |
 
 ### NVIDIA GPU
 
@@ -120,26 +119,25 @@ Same model memory as Apple Silicon, plus ~1 GB CUDA overhead.
 
 | VRAM | Quant | KV cache | Max context | Total VRAM used | Vision |
 |---|---|---|---|---:|---|
-| 16 GB | `IQ2_M` | `q4_0` | 150K | 15.8 GB | ✓ |
-| 16 GB | **`IQ3_M`** | `q4_0` | 90K | **15.9 GB** | ✗ |
-| 24 GB | **`Q4_K_M`** | `q4_0` | **262K** | **23.6 GB** | ✓ |
-| 24 GB | `IQ4_XS` | `q8_0` | 200K | 23.8 GB | ✓ |
-| 48 GB | **`Q6_K`** | `q8_0` | **262K** | **32.5 GB** | ✓ |
-| 48 GB | `Q8_0` | `q4_0` | 262K | 34.4 GB | ✓ |
-| 48 GB | `Q5_K_M` | `q4_0` | 262K | 25.6 GB | ✓ |
-| 80 GB | **`Q8_0`** | `q8_0` | **262K** | **38.4 GB** | ✓ |
+| 16 GB | **`IQ2_M`** | `q4_0` | **200K** | **15.7 GB** | ✓ |
+| 24 GB | **`Q4_K_M`** | `q4_0` | **262K** | **22.8 GB** | ✓ |
+| 24 GB | `Q5_K_M` | `q4_0` | 180K | 24.0 GB | ✓ |
+| 48 GB | **`Q6_K`** | `q8_0` | **262K** | **32.2 GB** | ✓ |
+| 48 GB | `Q8_0` | `q8_0` | 262K | 38.3 GB | ✓ |
 
 > **24 GB Mac:** `IQ3_M`/q4_0 — 128K with vision, 180K text-only.
 >
-> **32 GB Mac:** `Q5_K_M`/q4_0 — 262K text-only. For vision at 262K, use `Q4_K_M`. `Q5_K_M`/q8_0 for higher KV quality at 128K with vision.
+> **32 GB Mac:** `Q5_K_M`/q4_0 — 262K text-only. For vision at 262K, use `Q4_K_M`. `Q5_K_M`/q8_0 for higher KV quality at 128K text-only.
 >
-> **48 GB Mac:** `Q6_K`/q8_0 — best quality at 262K (30.4 GB). `Q8_0`/q8_0 for perfection (36.4 GB).
+> **48 GB+ Mac:** `Q6_K`/q8_0 — best quality at 262K with vision (31.2 GB). `Q8_0`/q8_0 for perfection (37.3 GB).
 >
-> **24 GB GPU:** `Q4_K_M`/q4_0 reaches 262K with vision. `IQ4_XS`/q8_0 for higher KV quality at 200K.
+> **16 GB GPU:** `IQ2_M`/q4_0 — 200K with vision.
 >
-> **48 GB GPU:** `Q6_K`/q8_0 — 262K at high quality (32.5 GB). `Q8_0`/q4_0 also fits (34.4 GB).
+> **24 GB GPU:** `Q4_K_M`/q4_0 reaches 262K with vision. `Q5_K_M`/q4_0 for higher quality at 180K with vision.
+>
+> **48 GB+ GPU:** `Q6_K`/q8_0 — 262K at high quality with vision (32.2 GB). `Q8_0`/q8_0 for perfection (38.3 GB).
 
-For coding and reasoning, prioritize higher quants with `q8_0` KV. For general chat and RAG, `IQ4_XS` with `q4_0` and larger context is often sufficient.
+For coding and reasoning, prioritize higher quants with `q8_0` KV. For general chat and RAG, lower quants with `q4_0` KV and larger context are often sufficient.
 
 Vision adds ~0.9 GB for mmproj. macOS needs **≥ 8 GB** for itself (16 GB Macs excepted — use ~4 GB). You can increase available memory by raising the wired memory limit, e.g. for a 96 GB Mac: `sudo sysctl iogpu.wired_limit_mb=90112` (88 GB). NVIDIA reserves ~1 GB for CUDA.
 
@@ -259,9 +257,9 @@ Effect on hardware requirements (Q5_K_M, 80K context):
 
 | KV type | Model + recurrent + KV | Hardware |
 |---|---|---|
-| f16 | 24 GB | 32 GB Mac |
+| f16 | 24 GB | 48 GB Mac |
 | **q8_0** | **22 GB** | **32 GB Mac** |
-| q4_0 | 20 GB | 24 GB Mac |
+| q4_0 | 21 GB | 32 GB Mac |
 
 ---
 
@@ -309,11 +307,11 @@ Uses cached n-grams from previous prompts.
 |---|---:|---|---|---|
 | `Qwen3.6-27B-Q8_0-mtp.gguf` | 27 GB | 48 GB Mac · 48 GB GPU | 48 GB Mac · 48 GB GPU | 48 GB Mac · 48 GB GPU |
 | `Qwen3.6-27B-Q6_K-mtp.gguf` | 21 GB | 32 GB Mac · 24 GB GPU | 48 GB Mac · 48 GB GPU | 48 GB Mac · 48 GB GPU |
-| **`Qwen3.6-27B-Q5_K_M-mtp.gguf`** | **18 GB** | **32 GB Mac · 24 GB GPU** | **32 GB Mac · 24 GB GPU** | **32 GB Mac · 24 GB GPU** |
+| **`Qwen3.6-27B-Q5_K_M-mtp.gguf`** | **18 GB** | **32 GB Mac · 24 GB GPU** | **32 GB Mac · 24 GB GPU** | **32 GB Mac · 48 GB GPU** |
 | `Qwen3.6-27B-Q4_K_M-mtp.gguf` | 16 GB | 32 GB Mac · 24 GB GPU | 32 GB Mac · 24 GB GPU | 32 GB Mac · 24 GB GPU |
-| `Qwen3.6-27B-IQ4_XS-mtp.gguf` | 14 GB | 24 GB Mac · 16 GB GPU | 32 GB Mac · 24 GB GPU | 32 GB Mac · 24 GB GPU |
+| `Qwen3.6-27B-IQ4_XS-mtp.gguf` | 14 GB | 24 GB Mac · 24 GB GPU | 32 GB Mac · 24 GB GPU | 32 GB Mac · 24 GB GPU |
 | `Qwen3.6-27B-IQ3_M-mtp.gguf` | 12 GB | 24 GB Mac · 16 GB GPU | 24 GB Mac · 24 GB GPU | 32 GB Mac · 24 GB GPU |
-| `Qwen3.6-27B-IQ2_M-mtp.gguf` | 10 GB | 16 GB Mac · 16 GB GPU | 24 GB Mac · 16 GB GPU | 24 GB Mac · 24 GB GPU |
+| `Qwen3.6-27B-IQ2_M-mtp.gguf` | 10 GB | 16 GB Mac · 16 GB GPU | 24 GB Mac · 16 GB GPU | 24 GB Mac · 16 GB GPU |
 | `mmproj-Qwen3.6-27B-f16.gguf` | 885 MB | Vision encoder (optional, any tier) | — | — |
 
 All tiers include MTP heads. Q8_0 is the direct conversion; all other tiers were quantized from it with an importance matrix. Q5_K_M is the sweet spot — use Q4_K_M if you're tight on RAM, Q8_0 if you want perfection. GPU means NVIDIA (RTX 3090/4090 = 24 GB, A6000 = 48 GB, A100 = 80 GB).