{"v":1,"id":"model:hf:KyleHessling1/Qwopus3.6-27B-Fusion-GGUF","slug":"model-kylehessling1-qwopus3-6-27b-fusion-gguf","kind":"model","category":"llm","title":"Qwopus3.6-27B-Fusion-GGUF","summary":"▶ PLAY ORBITAL — a game this model built, live in your browser ◀","source":{"provider":"hf","ref":"KyleHessling1/Qwopus3.6-27B-Fusion-GGUF","url":"https://huggingface.co/KyleHessling1/Qwopus3.6-27B-Fusion-GGUF","rev":"a622cd044d67247cd6a66df66753f6a0a8511b78","fetchedAt":"2026-10-02T21:01:10.613Z","etag":"W/\"2f2f-/JxJeOLwtzHrcgwdpUiX4NXasWw\""},"author":{"name":"KyleHessling1","url":"https://huggingface.co/KyleHessling1"},"license":{"spdx":null,"raw":"other","url":"https://huggingface.co/Qwen/Qwen3.6-27B/blob/main/LICENSE","open":null,"note":"custom license: read it at the source before installing"},"metrics":{"downloads":291886,"downloadsWeek":327838,"likes":78,"takenAt":"2026-10-02T21:01:10.613Z"},"tags":["gguf","merge","task-vector","dare-ties","layer-weighted","qwen3","qwen35","reasoning","code","llama.cpp","experimental","text-generation","en","endpoints_compatible","conversational"],"pipeline":"text-generation","links":{"github":"ggml-org/llama.cpp","npm":"node-llama-cpp"},"updatedAt":"2026-08-04T18:02:18.000Z","collectedAt":"2026-10-02T21:01:10.613Z","review":{"numbers":["291,886 downloads on Hugging Face","78 likes","license other","16 GB for Qwopus3.6-27B-Fusion-Q4_K_M.gguf","327,838 npm downloads a week for node-llama-cpp","latest node-llama-cpp@3.22.1"],"log":null},"trust":"unlabeled","health":{"status":"alive","checkedAt":"2026-10-02T21:01:10.613Z","http":200},"description":"▶ PLAY ORBITAL — a game this model built, live in your browser ◀\n\n> **Research preview.** This model attempts to combine the **reasoning** capability of\n> `Qwopus3.6-27B-v2` with the **code-execution** capability of `Qwopus3.6-27B-Coder` — in a single 27B\n> model, *without significant loss to either* skill. In practice it behaves like a production model\n> with both capabilities fused, but it has **not** yet been through a full rigorous evaluation. If you\n> find issues, please reach out on X — **[@kylehessling1](https://x.com/kylehessling1)**.\n\n- **Format:** GGUF, `Q4_K_M` (~16.5 GB) — runs on a single 24–32 GB GPU via `llama.cpp`.\n- **Architecture:** `qwen35` hybrid (linear-attention + periodic full-attention), 27B params, 64 layers.\n- **Native context:** 262,144 tokens. Verified clean (needle + termination) to 60K.\n- **Modes:** thinking (reasoning) **on** by default; strong agentic/coding behavior.\n- **Vision:** 🖼️ **image input supported** via the companion projector `Qwopus3.6-27B-Fusion-mmproj-Q8_0.gguf` (~0.6 GB) — pair it with *any* quant below (see [Vision](#-vision-multimodal)).\n\n━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\n\n## ◆ TL;DR\n\n| | |\n|---|---|\n| **Base / anchor** | `Qwen/Qwen3.6-27B` (both parents descend from it) |\n| **Parent A — reasoning** | `Qwopus3.6-27B-v2` — donates the merge base + embeddings/head/norms/MTP |\n| **Parent B — code** | `Qwopus3.6-27B-Coder` — delta source, injected with depth-increasing weight |\n| **Merge math** | `W(L) = V2 + α(L)·(Coder − V2)`, α linear ramp **0.12 (early) → 0.48 (late)** |\n| **Frozen (copied from A)** | `embed_tokens`, `lm_head`, all `norm`, `mtp`/NextN, any vision tensors |\n\n**Headline results** (Q4_K_M, thinking-on unless noted):\n\n| Capability | Score | Stability | Score |\n|---|---|---|---|\n| HumanEval | **94.5%** | Termination @ temp 0.2 & 0.9 | **4/4 clean** |\n| MBPP | **87.9%** | Temp sweep 0.5…\n\nSource: https://huggingface.co/KyleHessling1/Qwopus3.6-27B-Fusion-GGUF","install":{"kind":"model","hfId":"KyleHessling1/Qwopus3.6-27B-Fusion-GGUF","gated":false,"format":"gguf","files":[{"name":"Qwopus3.6-27B-Fusion-BF16.gguf","size":54657731680,"quant":"BF16","sha256":"541eb4a82bcaf4084b2d230e860080acc78924faab630a9e8877af3001f6ab68"},{"name":"Qwopus3.6-27B-Fusion-Q3_K_M.gguf","size":13500734560,"quant":"Q3_K_M","sha256":"6b924e8f4e2be94a69e2e9d30c1b86f06856538466c4c6abd82ad4cb05abf2c2"},{"name":"Qwopus3.6-27B-Fusion-Q4_K_M.gguf","size":16810712160,"quant":"Q4_K_M","sha256":"6d4e6e9ffba84ffe5166969303d4918f3a31682d75862416827412cee8f20d5d"},{"name":"Qwopus3.6-27B-Fusion-Q5_K_M.gguf","size":19535699040,"quant":"Q5_K_M","sha256":"eeb9b184a180136b21fdd58e1cc66a47e031c176f7c1a76234ee469ce776609c"},{"name":"Qwopus3.6-27B-Fusion-Q6_K.gguf","size":22430997600,"quant":"Q6_K","sha256":"44007c09e95a967e476703a629a64fb328ea327a3470ecae54db83ff8f3973bb"},{"name":"Qwopus3.6-27B-Fusion-Q8_0.gguf","size":29047082080,"quant":"Q8_0","sha256":"5594e1776b75beedf4a54b933bba386dc83a0883417e3bcd9ef53fdfd120d5b6"},{"name":"Qwopus3.6-27B-Fusion-mmproj-Q8_0.gguf","size":629246880,"quant":"Q8_0","sha256":"eb561d41a7bbeb0fcf04883c8af11078ef6cae0a66862a0b68443cfca495269d"}],"totalBytes":156612204000,"suggestedFile":"Qwopus3.6-27B-Fusion-Q4_K_M.gguf","requirements":{"ramGb":19,"diskBytes":16810712160,"note":"estimate: suggested file size × 1.15 + 0.5 GB; a real measurement comes with lsh models"},"runWith":["llama.cpp","ollama"],"command":"lsh models install hf:KyleHessling1/Qwopus3.6-27B-Fusion-GGUF"}}