Downloads · 30 days
510
32% of all-time downloads
sigmanih/Qwen-Qwen2.5-Coder-14B-Instruct-GGUF-Q6_K
Qwen-Qwen2.5-Coder-14B-Instruct-GGUF-Q6_K is a text generation model from sigmanih. Use it when you need the model to write or continue text. The card lists the license as other.
Downloads · 30 days
510
32% of all-time downloads
All-time downloads
1.6K
Public
Repo size
12.1 GB
Likes
1
Public
Click a slice to open those files.
.gguf12.1 GB · 100%
From the Hugging Face model README
❤️ Support & Community: If you find this model helpful, please give this repository a Like on Hugging Face and a ⭐ Star on our SigmaStudio GitHub!
Qwen-Qwen2.5-Coder-14B-Instruct-GGUF-Q6_K is a production-ready model optimized and published using the Model Hub module of Sigma Studio.
| Specification | Value |
|---|---|
| Model Repository | sigmanih/Qwen-Qwen2.5-Coder-14B-Instruct-GGUF-Q6_K |
| Weight Format | Sconosciuto (sconosciuta) |
| Base Architecture | CausalLM |
| Active Parameters | 14B |
| Context Window | 32,768 tokens |
| Total Disk Footprint | 0.0 GB |
| Inference RAM / VRAM | ~12.4 GB VRAM (Full GPU offload) / ~11.5 GB RAM (CPU/Hybrid) |
| Recommended Hardware | GPU with 12-16 GB VRAM (e.g. RTX 3060/4070 or 32 GB RAM) |
| Recommended Usage | High-speed coding, Everyday assistants, Autonomous agentic loops & reasoning. |
Evaluated directly on GPU via Sigma Studio Training Lab (Deterministic seed 42, Temp 0.0):
| Benchmark Suite | Score / Accuracy | Total Questions Evaluated | Pass Rate | Test Date | Execution Engine |
|---|---|---|---|---|---|
| Tutti i Benchmark Ufficiali | 78.0% | 78/100 quesiti superati | 78.0% Pass | 2026-09-02 | ⚡ SigmaEngine Direct GPU |
Side-by-side performance comparison between direct zero-overhead answer (No-Thinking) and Chain-of-Thought step-by-step reasoning (Deep Thinking CoT):
⚡ No-Thinking (Direct Response) : [█████████████████░░░] 84.2% (48/57 passed)
🧠 Deep Thinking (CoT Reasoning) : [██████████████░░░░░░] 69.8% (30/43 passed)
📈 CoT Performance Delta : -14.4%
%%{init: {'theme': 'dark'}}%%
xychart-beta
title "Accuracy (%): No-Thinking vs Deep Thinking"
x-axis ["⚡ No-Thinking (Direct)", "🧠 Deep Thinking (CoT)"]
y-axis "Accuracy (%)" 0 --> 100
bar [84.2, 69.8]
| Execution Mode | Accuracy (%) | Questions Passed | Operational Profile & Latency |
|---|---|---|---|
| ⚡ No-Thinking (Direct Response) | 84.2% | 48 / 57 | Minimal latency, immediate token-to-first-byte, zero reasoning tokens overhead |
| 🧠 Deep Thinking (CoT Reasoning) | 69.8% | 30 / 43 | Multi-step structured reasoning trace (-14.4%), optimal for math & hard logic |
| Dataset / Benchmark Suite | Domain / Category | Correct / Total | Accuracy (%) | Status |
|---|---|---|---|---|
| ARC-Challenge | Science & Grade-School Reasoning | 9 / 9 | 100% | ✅ Passed |
| BIG-Bench Hard | Complex Multi-Task Logic & Symbolics | 7 / 7 | 100% | ✅ Passed |
| GPQA | Graduate-Level Academic Reasoning | 2 / 9 | 22% | ⚠️ Low |
| GSM8K | Multi-Step Grade School Math | 8 / 9 | 89% | ✅ Passed |
| HellaSwag | Commonsense Reasoning & Situational NLI | 7 / 9 | 78% | ✅ Passed |
| HumanEval | Python Coding (pass@1) | 7 / 7 | 100% | ✅ Passed |
| MATH | Championship Competition Math | 8 / 9 | 89% | ✅ Passed |
| MBPP | Python Programming with Unit Tests | 9 / 9 | 100% | ✅ Passed |
| MMLU | General Knowledge & Multi-Subject | 8 / 14 | 57% | ⚡ Fair |
| MMLU-Pro | Advanced Multi-Step Reasoning | 5 / 9 | 56% | ⚡ Fair |
| TruthfulQA | Factuality & Anti-Hallucination | 8 / 9 | 89% | ✅ Passed |
| 🏆 OVERALL TOTAL | All Evaluated Datasets | 78 / 100 | 78% | 🏆 78% Pass |
Protocol: code_execution, continuation_logprob, cot_generation, letter_logprob · temp 0.0 · seed 42
Reproducibility hash: SHA256-28372CB2003770FC
⚠️ Measured on a slice of the dataset, not the full suite: this score is not comparable with a full-suite run.
Empirical multi-turn agent reliability evaluation (zero quizzes, fully grounded file-system actions inside isolated Sandbox Jail):
| Protocol Metric | Outcome | Validation Criteria & Details |
|---|---|---|
| Tool Protocol Adherence | 77.2% | 71/92 analytical criteria verified |
| Autonomous Task Completion | 37.5% | 3/8 scenarios completed with exact target file |
| Sandbox Jail Containment | 100% Compliant | Zero escape attempts outside workspace boundary |
| Execution Efficiency | 90 turns (353.7s) | Optimal multi-step turn and token budget usage |
| Scenario Name | Difficulty Tier | Score | Turns Used | Goal Status |
|---|---|---|---|---|
| File Nuovo | Livello 1 (Base) | 91% | 6/14 | ✅ Passed |
| Ispezione Chiave | Livello 1 (Base) | 64% | 14/14 | ⚠️ Partial |
| Modifica Mirata | Livello 2 (Intermedio) | 64% | 14/14 | ⚠️ Partial |
| Ricerca Albero | Livello 2 (Intermedio) | 73% | 14/14 | ⚠️ Partial |
| Debug E Fix | Livello 3 (Avanzato) | 73% | 14/14 | ⚠️ Partial |
| Refactor Due File | Livello 3 (Avanzato) | 100% | 8/14 | ✅ Passed |
| Si Ripara Da Solo | Livello 4 (Autonomia) | 77% | 14/14 | ⚠️ Partial |
| Premessa Sbagliata | Livello 4 (Autonomia) | 77% | 6/14 | ✅ Passed |
Measured on NVIDIA GeForce RTX 5070 Ti • 15.9 GB VRAM. Two different numbers follow, and they are not interchangeable.
| What was measured | Value | How |
|---|---|---|
| Single-stream decode (what a chat feels) | 34.8 tok/s | one request at a time, on NVIDIA GeForce RTX 5070 Ti |
| Prompt processing | 137 tok/s | same probe |
| Aggregate throughput during evaluation | 34.4 tok/s | several requests in flight — not what a single answer runs at |
Speeds on other hardware were not measured and are not guessed here. A single-stream figure from one machine cannot be scaled into a prediction for another: it depends on memory bandwidth, quantization, context length and driver, and the error is large enough to be misleading.
Launch Sigma Studio to enjoy full 1-click GPU hardware acceleration, live monitoring, and visual chat:
# Clone and run Sigma Studio
git clone https://github.com/Sigmanih/SigmaStudio.git
cd SigmaStudio
.\sigma_studio.bat
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
model_id = "sigmanih/Qwen-Qwen2.5-Coder-14B-Instruct-GGUF-Q6_K"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16, device_map="auto")
Qwen-Qwen2.5-Coder-14B-Instruct-GGUF-Q6_K è un modello ottimizzato pronto per l'inferenza e l'integrazione locale, pubblicato attraverso Σ-SIGMA Studio.
CausalLM (14B parametri)Sconosciuto (sconosciuta)0.0 GB~12.4 GB VRAM (offload GPU completo) | ~11.5 GB RAM (inferenza CPU/ibrida)32,768 tokenTutti i Benchmark Ufficiali78.0% (78/100 quesiti superati)Confronto visuale tra risposta istantanea diretta (No-Thinking) e ragionamento guidato multi-step (Deep Thinking CoT):
⚡ No-Thinking (Risposta Diretta) : [█████████████████░░░] 84.2% (48/57 superati)
🧠 Deep Thinking (CoT Reasoning) : [██████████████░░░░░░] 69.8% (30/43 superati)
📈 Delta Prestazionale CoT : -14.4%
%%{init: {'theme': 'dark'}}%%
xychart-beta
title "Accuratezza (%): No-Thinking vs Deep Thinking"
x-axis ["⚡ No-Thinking (Diretto)", "🧠 Deep Thinking (CoT)"]
y-axis "Accuratezza (%)" 0 --> 100
bar [84.2, 69.8]
| Modalità di Esecuzione | Accuratezza (%) | Quesiti Superati | Profilo Operativo & Latenza |
|---|---|---|---|
| ⚡ No-Thinking (Risposta Diretta) | 84.2% | 48 / 57 | Latenza minima, token-to-first-byte istantaneo, zero overhead di ragionamento |
| 🧠 Deep Thinking (CoT Reasoning) | 69.8% | 30 / 43 | Risoluzione passo-passo multi-step (-14.4%), ideale per logica complessa e matematica |
| Dataset / Suite di Test | Ambito / Dominio | Corretti / Totale | Accuratezza (%) | Esito |
|---|---|---|---|---|
| ARC-Challenge | Ragionamento Scientifico Avanzato | 9 / 9 | 100% | ✅ Superato |
| BIG-Bench Hard | Logica Complessa & Compiti Multi-Fase | 7 / 7 | 100% | ✅ Superato |
| GPQA | Ragionamento Accademico di Livello Laurea | 2 / 9 | 22% | ⚠️ Migliorabile |
| GSM8K | Matematica & Logica Multi-Step | 8 / 9 | 89% | ✅ Superato |
| HellaSwag | Buon Senso & Comprensione Situazionale | 7 / 9 | 78% | ✅ Superato |
| HumanEval | Sintesi Codice Python (pass@1) | 7 / 7 | 100% | ✅ Superato |
| MATH | Matematica Olimpica & Competitiva | 8 / 9 | 89% | ✅ Superato |
| MBPP | Programmazione Python con Unit Test | 9 / 9 | 100% | ✅ Superato |
| MMLU | Conoscenza Generale Multidisciplinare | 8 / 14 | 57% | ⚡ Discreto |
| MMLU-Pro | Ragionamento Avanzato Multi-Step | 5 / 9 | 56% | ⚡ Discreto |
| TruthfulQA | Fattualità & Resistenza ad Allucinazioni | 8 / 9 | 89% | ✅ Superato |
| 🏆 TOTALE COMPLESSIVO | Tutti i Dataset Valutati | 78 / 100 | 78% | 🏆 78% Pass |
Protocollo: code_execution, continuation_logprob, cot_generation, letter_logprob · temp 0.0 · seed 42
Impronta di riproducibilità: SHA256-28372CB2003770FC
⚠️ Misurato su una porzione del dataset, non sulla suite intera: il punteggio non è confrontabile con uno ottenuto sull'intero.
2026-09-02 su motore deterministico SigmaEngineValutazione empirica dell'affidabilità nei compiti di agente autonomo (nessun quiz teorico, solo esecuzioni e verifiche su filesystem isolato):
| Metrica di Protocollo | Risultato | Dettaglio e Criteri di Validazione |
|---|---|---|
| Aderenza Protocollo Tool | 77.2% | 71/92 prove analitiche verificate |
| Completamento Scenari Operativi | 37.5% | 3/8 scenari conclusi con output esatto |
| Sicurezza Sandbox Jail | 100% Conforme | Confinamento rigido, zero tentativi fuori sandbox |
| Efficienza Esecutiva | 90 turni (353.7s) | Rispetto rigoroso dei budget operativi per scenario |
| Scenario di Prova | Difficoltà | Punteggio | Turni | Obiettivo Verificato |
|---|---|---|---|---|
| File Nuovo | Livello 1 (Base) | 91% | 6/14 | ✅ Raggiunto |
| Ispezione Chiave | Livello 1 (Base) | 64% | 14/14 | ⚠️ Parziale |
| Modifica Mirata | Livello 2 (Intermedio) | 64% | 14/14 | ⚠️ Parziale |
| Ricerca Albero | Livello 2 (Intermedio) | 73% | 14/14 | ⚠️ Parziale |
| Debug E Fix | Livello 3 (Avanzato) | 73% | 14/14 | ⚠️ Parziale |
| Refactor Due File | Livello 3 (Avanzato) | 100% | 8/14 | ✅ Raggiunto |
| Si Ripara Da Solo | Livello 4 (Autonomia) | 77% | 14/14 | ⚠️ Parziale |
| Premessa Sbagliata | Livello 4 (Autonomia) | 77% | 6/14 | ✅ Raggiunto |
34.4 tok/s su NVIDIA GeForce RTX 5070 Ti.34.8 tok/s su NVIDIA GeForce RTX 5070 Ti.137 tok/s.34.4 tok/s — piu' richieste in volo insieme, non la velocita' di una risposta singola.Se questo modello ti è utile o vuoi esplorare l'ecosistema completo:
Creato e distribuito con il Model Hub di Σ-SIGMA Studio (03/11/2026 00:50)