GraySoft
Download guIDE →
Projects Models Compare Cloud benchmarks FAQ Download guIDE →
Model Intelligence Sheet

IsValorum/Qwen3.8-27B-EfficientThink-VAL-APEX-I-NanoPlus-GGUF overview

base model: nerkyor/Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2 base model relation: quantized quantized by: …

gguf27b27b-paramsllama.cppquantizedquantizationval-apex-iapexapex-quantapex-i-nanoplusnanoplusqwenqwen3.8reasoningchain-of-thoughtdflash2speculative-decodingsimpoimatrixtext-generationconversationalenbase_model:nerkyor/Qwen3.8-27B-Coder390-EfficientThink-Opus5.5-GPT6Astra-Grok4.7-DSV4Pro-K3-SFT-RLOO-MTP-DFlash2base_model:quantized:nerkyor/Qwen3.8-27B-Coder390-EfficientThink-Opus5.5-GPT6Astra-Grok4.7-DSV4Pro-K3-SFT-RLOO-MTP-DFlash2

Runs locally from ~498.1 MB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).

Downloads
7,555
Likes
4
Pipeline
text-generation
Author
IsValorum

Repository Files & Downloads

5 GGUF files detected
Direct downloads for local inference
FileTypeQuantizationSizeLink
Qwen3.8-27B-EfficientThink-VAL-APEX-I-NanoPlus-V3.ggufGGUFGGUF10.41 GBDownload
mmproj-Qwen3.8-27B-Q4_K_M.ggufGGUFQ4_K_M498.1 MBDownload
mmproj-Qwen3.8-27B-Q8_0.ggufGGUFQ8_0600.1 MBDownload
mtp-Qwen3.8-27B-Q4_0.ggufGGUFQ4_01.56 GBDownload
speculative/dflash2-qwen38-27b-Q4_K_M.ggufGGUFQ4_K_M1.06 GBDownload

Model Details

Model IDIsValorum/Qwen3.8-27B-EfficientThink-VAL-APEX-I-NanoPlus-GGUF
AuthorIsValorum
Pipelinetext-generation
Licenseapache-2.0
Base modelnerkyor/Qwen3.8-27B-Coder390-EfficientThink-Opus5.5-GPT6Astra-Grok4.7-DSV4Pro-K3-SFT-RLOO-MTP-DFlash2
Last modified2026-10-14T20:56:46.000Z

Model README

---

base_model: nerkyor/Qwen3.8-27B-Coder390-EfficientThink-Opus5.5-GPT6Astra-Grok4.7-DSV4Pro-K3-SFT-RLOO-MTP-DFlash2

base_model_relation: quantized

quantized_by: IsValorum

library_name: gguf

license: apache-2.0

language:

  • en

tags:

  • 27b
  • 27b-params
  • gguf
  • llama.cpp
  • quantized
  • quantization
  • val-apex-i
  • apex
  • apex-quant
  • apex-i-nanoplus
  • nanoplus
  • qwen
  • qwen3.8
  • reasoning
  • chain-of-thought
  • dflash2
  • speculative-decoding
  • simpo
  • imatrix

pipeline_tag: text-generation

---

Qwen3.8-27B-EfficientThink VAL-APEX-I NanoPlus V3 GGUF

>

> The legacy build suffered from severe recurrent state instability, context degradation, and unintended language mixing during generation.

>

> The current release completely resolves these issues:

> - Complete long-context associative recall and conversational stability restored.

> - Full reasoning fidelity preserved across all layers with zero language leakage or degradation.

> - Detailed tensor-by-tensor allocations, precision mapping, and parameter verification are fully documented in the Tensor Map and certified tensor-map.tsv file.

>

> Action Required: Please delete any older .gguf file downloaded prior to today and download the newly published certified build: Qwen3.8-27B-EfficientThink-VAL-APEX-I-NanoPlus-V3.gguf.

<a href="https://ko-fi.com/isvalorum"><img src="https://ko-fi.com/img/githubbutton_sm.svg" alt="Support on Ko-fi" height="38"></a>

Independently computed using dedicated cloud compute clusters, funded 100% out-of-pocket.

If you consider these releases to be of high quality and they help you in your day-to-day workflow, consider supporting on Ko-fi. Every contribution directly sustains future cloud compute hours and helps support my university studies.

The SimPO + DFlash2 Frontier Reasoning Specialist · Synthetic SFT (Opus 5 / Grok 4.6 / GPT 5.6 Sol) · Native 256K Context

Official VAL-APEX-I quantization of nerkyor/Qwen3.8-27B-Coder390-EfficientThink-Opus5.5-GPT6Astra-Grok4.7-DSV4Pro-K3-SFT-RLOO-MTP-DFlash2.

VAL-APEX-I stands for:

Vector-calibrated Asymmetric Layer-wise Outlier-preserving Recurrent-aware Unified Matrix-quantization

> [!NOTE]

> ### EXPLORE THE QWEN3.8 27B VAL-APEX-I EDITIONS

> Choose a model variant and quantization profile:

>

> - Swift-1.5 Qwen3.8 27B Uncensored MTP — MiniPlus V3 — 14.16 GB.

> - Swift-1.5 Qwen3.8 27B Uncensored MTP — NanoPlus V3 — 12.18 GB.

> - Qwen3.8 27B — MiniPlus V3 — 14.16 GB.

> - Qwen3.8 27B — NanoPlus V3 — 12.18 GB.

> - Qwen3.8 27B EfficientThink — MiniPlus V3 — 13.91 GB.

> - Qwen3.8 27B Huihui Abliterated — MiniPlus V3 — 14.16 GB.

> - Qwen3.8 27B Huihui Abliterated — NanoPlus V3 — 12.18 GB.

>

> Collections: MiniPlus Abliterated 27B · MiniPlus Standard 27B · NanoPlus Abliterated 27B · NanoPlus Standard 27B

> [!IMPORTANT]

> ### THE DEFINITIVE SPECIFICATION IN THE 11 to 12 GB CEILING

> This VAL-APEX-I NanoPlus V3 release is 11.93 GB, with 3.55 whole-file BPW and 851 audited tensors. Its reported WikiText-2 PPL is 5.6504 ± 0.4832.

---

<a id="quick-navigation"></a>Quick Navigation Index

  1. Quantization Metrics & Tensor Map
  2. Independent Community Benchmarks (NanoBench v0.4.0)
  3. Upstream Lineage & Ecosystem Compatibility (DFlash2, SGLang, vLLM, Lynn Agent)
  4. Model Files & Technical Specifications
  5. Native Context & Runtime Memory
  6. Recommended Configuration & Setup
  7. Recommended Generation Parameters
  8. CRITICAL: Coding Syntax & Repeat Penalty Advisory (Preventing Character Swapping)
  9. Hardened Agentic Chat Template & Reasoning Effort
  10. Community Compute Fund & Student Support

---

<a id="toc-comparison"></a>

<a id="toc-01"></a>

<a id="toc-02"></a>

<a id="toc-03"></a>

<a id="toc-05"></a>

1. Quantization Metrics

Size & Quality Metrics

| Edition / quantization | File size | Whole-file BPW | WikiText-2 PPL |

| :--- | :--- | :--- | :--- |

| Qwen3.8 27B EfficientThink — NanoPlus V3 (this release) | 11.93 GB<br>(11.12 GiB) | 3.55 | 5.6504 ± 0.4832 |

> [!NOTE]

> Independent Hardware & Compute Transparency:

> Generating multi-million token calibration matrices and running certified perplexity evaluations on local architectures requires dedicated high-end cloud nodes funded 100% out-of-pocket. If you rely on these MiniPlus/NanoPlus V3 builds for your local workloads, contributions to the IsValorum Compute Fund on Ko-fi directly sponsor the GPU hours required for future Day-0 handcrafted quantizations.

<a id="toc-nanobench"></a>

Independent Community Benchmarks (NanoBench v0.4.0)

Independent multi-harness empirical evaluation conducted by community member @userr99 on a dedicated local workstation (24 GB VRAM / 64 GB RAM using llama.cpp). Evaluated across the 12-task autonomous coding and agentic suite on NanoBench v0.4.0 (public discussion thread & test logs):

| Test Harness | Total Score | Solved Tasks | Agentic (Tool Use) | Coding (Implementation) | Systems (Integrity) | Avg. Runtime |

| :--- | :---: | :---: | :---: | :---: | :---: | :---: |

| Oh My Pi (OMP 18.7) | 96.7 / 100 | 10 / 12 | 100 / 100 | 100 / 100 | 86.7 / 100 | aprox. 1h 47m |

| Hermes Agent (v0.21.5) | 95.4 / 100 | 9 / 12 | 100 / 100 | 97.0 / 100 | 86.7 / 100 | aprox. 1h 47m |

| DeepSeek Harness (0.2.0-rc.2) | 90.4 / 100 | 9 / 12 | 100 / 100 | 97.0 / 100 | 66.7 / 100 | aprox. 1h 47m |

Key Architectural Insights:

  • Zero Tool-Calling Degradation: Achieved a unanimous 100/100 in Agentic tasks across all three harnesses, confirming that protecting output.weight in Q6_K and DeltaNet recurrent memory states (ssm_a in F32, ssm_beta in Q8_0) prevents JSON schema rejections and state drift over multi-step workflows.
  • Universal Harness Compatibility: Surpassed the 90+ score barrier across all three test harnesses without entering infinite retry loops or syntax divergence.

Tensor Precision Map

| Component | MiniPlus V3 | NanoPlus V3 | GSQ-RCO IQ3_S |

| :--- | :--- | :--- | :--- |

| Output head | Q6_K x1 | Q5_K x1 | Q4_K x1 |

| Token embeddings | Q4_K x1 | Q3_K x1 | IQ2_S x1 |

| Normalizations | F32 x161 | F32 x161 | F32 x161 |

| SSM A (ssm_a) | F32 x48 | F32 x48 | F32 x48 |

| SSM convolution (ssm_conv1d) | F32 x48 | F32 x48 | F32 x48 |

| SSM time-step bias (ssm_dt) | F32 x48 | F32 x48 | F32 x48 |

| SSM norm (ssm_norm) | F32 x48 | F32 x48 | F32 x48 |

| Attention gates (attn_gate) | Q4_0 x1<br>Q4_K x47 | Q4_0 x1<br>Q4_K x47 | IQ2_S x2<br>IQ3_S x18<br>IQ3_XXS x9<br>IQ4_XS x12<br>Q2_K x4<br>Q4_K x3 |

| Linear QKV (attn_qkv) | Q4_0 x1<br>Q4_K x47 | Q3_K x47<br>Q4_0 x1 | IQ2_XS x1<br>IQ2_XXS x1<br>IQ3_S x22<br>IQ3_XXS x13<br>IQ4_XS x9<br>Q2_K x1<br>Q4_K x1 |

| SSM alpha (ssm_alpha) | F32 x48 | F32 x48 | BF16 x48 |

| SSM beta (ssm_beta) | Q4_0 x1<br>Q8_0 x47 | Q4_0 x1<br>Q8_0 x47 | BF16 x48 |

| SSM output (ssm_out) | Q4_0 x1<br>Q5_K x47 | Q4_0 x1<br>Q4_K x47 | IQ3_S x22<br>IQ3_XXS x4<br>IQ4_XS x16<br>Q4_K x6 |

| Full attention Q (attn_q) | Q4_K x16 | Q3_K x16 | IQ2_XXS x1<br>IQ3_S x3<br>IQ3_XXS x3<br>IQ4_XS x2<br>Q2_K x6<br>Q4_K x1 |

| Full attention K (attn_k) | Q5_K x16 | Q4_K x16 | IQ2_S x1<br>IQ3_S x1<br>IQ3_XXS x1<br>IQ4_XS x8<br>Q4_K x5 |

| Full attention V (attn_v) | Q6_K x16 | Q5_K x16 | IQ3_S x6<br>IQ3_XXS x1<br>IQ4_XS x1<br>Q4_K x8 |

| Full attention output (attn_output) | Q6_K x16 | Q5_K x16 | IQ3_S x10<br>IQ3_XXS x1<br>IQ4_XS x2<br>Q4_K x3 |

| MLP down (ffn_down) | IQ4_XS x64 | IQ3_XXS x48<br>IQ4_XS x16 | IQ2_S x4<br>IQ2_XS x3<br>IQ3_S x22<br>IQ3_XXS x7<br>IQ4_XS x21<br>Q2_K x1<br>Q4_K x6 |

| MLP gate (ffn_gate) | IQ3_S x32<br>IQ3_XXS x32 | IQ3_XXS x64 | IQ1_M x1<br>IQ2_S x4<br>IQ2_XS x4<br>IQ2_XXS x1<br>IQ3_S x15<br>IQ3_XXS x21<br>IQ4_XS x15<br>Q2_K x1<br>Q4_K x2 |

| MLP up (ffn_up) | IQ3_S x32<br>IQ3_XXS x32 | IQ3_XXS x64 | IQ2_S x5<br>IQ2_XS x1<br>IQ2_XXS x2<br>IQ3_S x25<br>IQ3_XXS x18<br>IQ4_XS x10<br>Q4_K x3 |

x denotes tensor counts. All columns reflect exact GGUF tensor headers verified in tensor-map.tsv. Normalizations exclude separately listed SSM norms.

This release contains 851 audited tensors. The main language GGUF contains no blk.64.* tensors; MTP and DFlash2 draft models are provided as separate companion files.

V3 architectural allocation: NanoPlus V3 features an elevated 3-bit floor across all SwiGLU MLP blocks (ffn_gate and ffn_up set uniformly to IQ3_XXS across all 64 layers; ffn_down uses IQ4_XS in outer layers 0-7 / 56-63 and IQ3_XXS in core layers 8-55). Output head (output.weight) is calibrated to Q5_K (5.50 bpw) to maximize token throughput and preserve dynamic range, entirely eliminating the linguistic decay in non-English contexts associated with 2-bit sub-allocations.

Actual layer-0 formats: blk.0.attn_gate.weight, blk.0.attn_qkv.weight, blk.0.ssm_beta.weight and blk.0.ssm_out.weight use Q4_0. All 48 ssm_alpha tensors remain F32 for mathematical stability.

Complete tensor-by-tensor audit: names, formats, shapes and parameter counts.

---

<a id="toc-upstream"></a>

2. Upstream Lineage & Ecosystem Compatibility

This model is derived from nerkyor/Qwen3.8-27B-Coder390-EfficientThink-Opus5.5-GPT6Astra-Grok4.7-DSV4Pro-K3-SFT-RLOO-MTP-DFlash2 and inherits its full ecosystem compatibility:

1. Training Lineage & Data Mixture

  • Pretrained Base: Qwen3.8-27B hybrid linear-quadratic architecture.
  • Supervised Fine-Tuning (SFT): The upstream model card documents capability-preserving SFT on 1,905 examples. The release identifier names K3 / Opus 5 / Grok 4.6 / GPT 5.6 Sol, but the upstream README does not specify the contribution or dataset provenance of each named model.
  • Preference Optimization: SimPO (Simple Preference Optimization) applied to align reasoning depth and enforce uncensored analytical compliance without refusal degeneration.
  • Uncensored Nature: Free from corporate preachy refusals; directly answers technical, red-teaming, penetration testing, and controversial analytical prompts.

2. DFlash2 Multi-Token Prediction (MTP) Speculative Decoding

  • Separate Draft Models: This repository provides companion MTP and DFlash2 GGUF files. The audited main language GGUF contains no blk.64.* tensors; the draft models are separate files.
  • Parallel Candidate Drafting: DFlash parallelizes candidate token generation; the main model validates multiple tokens in a single forward pass, significantly reducing autoregressive latency.
  • Compatibility with llama.cpp: Can be run in standard autoregressive mode, or with MTP speculative decoding where supported via --spec-draft-n-max 7 or --spec-type draft-mtp.

3. Serving Engine Support

  • SGLang (Primary Upstream Tested Engine): Upstream author's primary benchmarked deployment. Supports SGLang DFlash2 verify block size 8.
  • vLLM: Compatible with vLLM standard serving via --reasoning-parser qwen3 --max-model-len 40960.
  • Lynn Agent (v0.87.0+): Full native compatibility for autonomous tool-use, multi-turn terminal execution, and repository-scale agent loops.
  • llama.cpp / llama-server: 100% plug-and-play with official llama.cpp builds (b11000+) supporting hybrid DeltaNet SSM architectures.

---

<a id="toc-04"></a>

3. Model Files & Technical Specifications

| File Name | File Size | Weight file size | BPW / Type | Description |

| :--- | :--- | :--- | :--- | :--- |

| Qwen3.8-27B-EfficientThink-VAL-APEX-I-NanoPlus Qwen3.8-27B-EfficientThink-VAL-APEX-I-NanoPlus-V3.gguf | 11.93 GB (11.12 GiB) | 11.12 GiB | 3.55 BPW | Core language, uncensored frontier reasoning, CoT thought blocks & hybrid SSM/attention |

| mmproj-Qwen3.8-27B-Q4_K_M.gguf | 522.29 MB | 522.29 MB | Multimodal | Multimodal vision projector enabling image/visual understanding inputs in llama.cpp |

| mtp-Qwen3.8-27B-Q4_0.gguf | 1.68 GB | 1.68 GB | Draft MTP | Multi-Token Prediction draft adapter for accelerated speculative decoding |

| dflash2-qwen38-27b-Q4_K_M.gguf | 1.14 GB | 1.14 GB | Draft DFlash2 | High-speed DFlash2 speculative decoding draft model |

  • Base Model: nerkyor/Qwen3.8-27B-Coder390-EfficientThink-Opus5.5-GPT6Astra-Grok4.7-DSV4Pro-K3-SFT-RLOO-MTP-DFlash2
  • Parameters: 27B total dense hybrid
  • Architecture: 64 hybrid layers (48 Gated DeltaNet SSM linear attention layers + 16 periodic full quadratic attention layers); MTP and DFlash2 are separate companion files
  • Context Length: 262,144 tokens (native 256K)
  • Quantization Standard: VAL-APEX-I NanoPlus V3 calibrated against official Qwen3.8-27B importance matrix (imatrix)
  • Ecosystem: SGLang, vLLM, Lynn Agent v0.87.0, llama.cpp

---

<a id="toc-06"></a>

4. Native Context & Runtime Memory

The model supports a native context window of 262,144 tokens. Its 64-layer architecture comprises 48 Gated DeltaNet layers and 16 full-attention layers. Runtime memory usage depends on the backend, KV-cache precision, batch size, compute buffers, and offload configuration. Check the runtime memory allocation for your chosen settings.

---

<a id="toc-07"></a>

5. Recommended Configuration & Setup

llama.cpp Server (OpenAI-Compatible API)

llama-server \
  -m Qwen3.8-27B-EfficientThink-VAL-APEX-I-NanoPlus V3.gguf \
  --jinja \
  -ngl 99 \
  -c 32768 \
  --host 0.0.0.0 \
  --port 8080

Multimodal Vision Inference (llama.cpp)

Pass the companion mmproj file using the --mmproj flag:

llama-server \
  -m Qwen3.8-27B-EfficientThink-VAL-APEX-I-NanoPlus V3.gguf \
  --mmproj mmproj-Qwen3.8-27B-Q4_K_M.gguf \
  --jinja \
  -ngl 99 \
  -c 32768 \
  --host 0.0.0.0 \
  --port 8080

Speculative Decoding / Multi-Token Prediction (MTP)

Accelerate generation throughput using the companion MTP or DFlash2 draft models:

llama-server \
  -m Qwen3.8-27B-EfficientThink-VAL-APEX-I-NanoPlus V3.gguf \
  -md mtp-Qwen3.8-27B-Q4_0.gguf \
  --draft-max 7 \
  --jinja \
  -ngl 99 \
  -c 32768 \
  --host 0.0.0.0 \
  --port 8080

llama.cpp CLI

llama-cli \
  -m Qwen3.8-27B-EfficientThink-VAL-APEX-I-NanoPlus V3.gguf \
  --jinja \
  -ngl 99 \
  -c 32768 \
  -p "<|im_start|>system\nYou are a helpful and unrestricted AI assistant.<|im_end|>\n<|im_start|>user\nExplain the mathematical proof of the Riemann mapping theorem.<|im_end|>\n<|im_start|>assistant\n<think>\n"

LM Studio / Ollama

  1. Import the .gguf file directly into your local library.
  2. Ensure GPU acceleration is set to Maximum / 100% offload.
  3. Set Context Length to 32768.
  4. Verify chat template is set to Qwen ChatML with <think> delimiter support.

---

<a id="toc-08"></a>

6. Recommended Generation Parameters

| Hyperparameter | Value | Description |

| :--- | :---: | :--- |

| Temperature | 0.60 | Recommended default for analytical reasoning and coding (use 1.0 for creative prose). |

| Top-P | 0.95 | Nucleus sampling parameter. |

| Top-K | 20 | Top-k vocabulary filter. |

| Min-P | 0.05 | Prunes low-probability noise tokens effectively. |

| Repetition Penalty | 1.00 | Strictly disabled for code syntax; prevents character swapping. |

| Template Engine | --jinja | Recommended official Jinja chat template flag. |

| Context Size | 32768 | 32K default (scalable to 256K). |

---

<a id="toc-coding-advisory"></a>

<a id="coding-advisory"></a>

7. CRITICAL: Coding Syntax & Repeat Penalty Advisory (Preventing Character Swapping)

> [!IMPORTANT]

> ### PREVENTING SYNTAX & TOKEN SWAPPING IN CODE WORKFLOWS

> In programming code, brackets ({, }), assignment operators (=), and indentation whitespace repeat constantly across multi-line structures.

>

> Common Issue: Many local frontends (such as LM Studio defaults, Ollama, or web interfaces) ship with repeat_penalty set to 1.1 or 1.15. While this prevents loops in creative prose, applying repeat penalties to code artificially penalizes necessary syntax tokens. When the logit of { drops, the model is forced to emit the next closest mathematical token (= or [), resulting in character swapping or dropped/doubled whitespace.

>

> Eliminating Character Swapping:

> 1. Disable Repeat Penalties (Required for Code):

> - repeat_penalty: 1.0 (strictly disabled)

> - presence_penalty: 0.0

> - frequency_penalty: 0.0

> 2. Calibrate Samplers:

> - temperature: 0.60 (or 0.20 - 0.30 for strict, deterministic code syntax)

> - min_p: 0.05 (prunes low-probability noise tokens effectively)

> - top_p: 0.95

> - top_k: 20

> 3. Native Jinja Formatting: Always pass the --jinja flag so the tokenizer handles leading-space BPE tokens cleanly.

---

<a id="toc-chat-template"></a>

<a id="chat-template"></a>

8. Hardened Agentic Chat Template & Reasoning Effort

> [!TIP]

> ### MULTI-LEVEL REASONING EFFORT CONTROL

> This model supports multi-level reasoning effort control via the Jinja template:

> - low / minimal: Keeps internal thinking concise and focused strictly on immediate execution steps to minimize latency in automated loops.

> - medium (default): Balanced, structured reasoning process with standard analytical depth.

> - high / xhigh: Guides the model to formulate a clear implementation plan upfront before generating response, avoiding circular self-doubt loops.

> - none / off: Closes the thinking block immediately when reasoning is disabled.

---

<a id="toc-09"></a>

10. Community Compute Fund & Student Support

<a href="https://ko-fi.com/isvalorum"><img src="https://huggingface.co/spaces/IsValorum/MiniPlus-NanoPlus-Requests/resolve/main/assets/dance-gold-ship.gif" alt="Gold Ship dancing" width="128" align="right"></a>

All IsValorum quantizations will always remain completely free and open to the public without paywalls.

However, cloud GPU compute is expensive and funded entirely out-of-pocket. If you find these handcrafted builds to be of high quality and they benefit your day-to-day workflow, you can support future Day-0 releases on Ko-fi. Every contribution directly funds the cloud compute hours required for new model runs and helps support my university education:

<a href="https://ko-fi.com/isvalorum"><img src="https://ko-fi.com/img/githubbutton_sm.svg" alt="Support on Ko-fi" height="38"></a>

(When supporting on Ko-fi, feel free to leave a note with your Hugging Face handle and the specific model you would like prioritized).

<div style="clear: both;"></div>

Run IsValorum/Qwen3.8-27B-EfficientThink-VAL-APEX-I-NanoPlus-GGUF with guIDE

Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.

Download guIDE → · Browse 524k+ models · Compare models

Source: Hugging Face · Compare models