Downloads · 30 days
55
7% of all-time downloads
FahadPrimeX-b27-V3-GGUF
FahadPrimeX/FahadPrimeX-b27-V3-GGUF
FahadPrimeX-b27-V3-GGUF is a machine learning model from FahadPrimeX. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Quantized (GGUF) builds of FahadPrimeX-b27-V3: Q4KM (16.5GB) · Q5KM (19.2GB) · Q6K (22.1GB).
All-time downloads
835
Public
Repo size
116 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.gguf57.9 GB · 100%
At a glance
- License
- apache-2.0
- Access
- Public
- Created
- Sep 21, 2026
- Updated
- Sep 21, 2026
- SHA
- 9b314714
Model card
From the Hugging Face model README
FahadPrimeX-b27-V3-GGUF ⚡
Quantized (GGUF) builds of FahadPrimeX-b27-V3: Q4_K_M (16.5GB) · Q5_K_M (19.2GB) · Q6_K (22.1GB).
Run locally (GGUF)
Works with llama.cpp / Ollama / LM Studio:
ollama run hf.co/FahadPrimeX/FahadPrimeX-b27-V3-GGUF:FahadPrimeX-b27-V3-Q4_K_M
Identity (all languages): "I am FahadPrimeX (b27), a Three.js-specialist model" / «أنا فهد برايم إكس (FahadPrimeX-b27) — نموذج متخصص في Three.js» — no family name. The model truthfully credits its Qwen3.8-27B lineage when asked who developed it.
Measured results (honest, 32 held-out tasks, same browser-validator & protocol)
| Model | Overall | Hard gates (render/controls/one-click) | Identity (3 languages) |
|---|---|---|---|
| Pre-RLVR base (V2) | 54.5% | 8/32 | 0/10 (says "Qwen") |
| V3 (this model) | 50.9% | 4/32 | 10/10 ✅ |
Read this honestly: V3 does not yet improve game-generation quality over its starting point — it does fully fix trilingual identity. An improved run (V3.1) with hard-gate-weighted rewards from round 0 is planned.
Architecture (inherited from Qwen3.8-27B)
A hybrid 64-layer design: 48 linear-attention layers (DeltaNet-style, constant-memory recurrence) interleaved with 16 full-attention layers (every 4th layer), plus one MTP/nextn draft layer used only by the original spec (excluded from our GGUF exports via --no-nextn so block_count matches the real 64 blocks).
- License
- apache-2.0
- Languages
- ar, en, zh
- Created
- Sep 21, 2026
- Updated
- Sep 21, 2026