qwen-2.5-14B-instruct-1m-gguf-F16

Qwen2.5-14B-Instruct-1M · GGUF F16

<!-- pbh-site-link -->

Evaluation results: this F16 build is the baseline each quantized build of this model is measured against. Its pass rates by capability family, with confidence intervals, are on the results page: pbhappliedsystems.com/models/qwen2.5-14b-instruct-1m#F16

<!-- pbh-card4 -->

How much does quantization degrade Qwen2.5-14B-Instruct-1M?

Answer (quant_eval v7.22): this F16 build is the baseline. Against it:

• Q4_K_M degraded 1 of 8 families, improved 3 — Q4_K_M results

Licence provenance: Apache-2.0 (SPDX Apache-2.0), commercial use permitted_by_recorded_license; upstream Qwen/Qwen2.5-14B-Instruct-1M at revision `` (unavailable_for_pre_provenance_legacy_artifact); evaluated file SHA-256 de08ea9c41234ef83b7aacf07f9ebc3cbaa20ca8aeb5f6417758a8798660aaa9.

Other builds of this model: Q4_K_M · All evaluated builds (Collection) · Datasets

Contents: Model Description · PBH Applied Systems Evaluation — quant_eval v7.22.22 · Comparability & Fixture Methodology · Per-Family Pass Rates — Full-Weight Baseline · Key Findings · Quantization Variants Comparison · Artifact Provenance · Evaluation Methodology

<!-- /pbh-card4 -->

Converted and evaluated by PBH Applied Systems, LLC — Applied AI/ML Consulting · LLM Optimization & Deployment · Quantized AI Infrastructure

🔬 This repository is part of a production-oriented evaluation series. Every model published under pbhappliedsystems has been independently evaluated using quant_eval v7.22.21/v7.22.22 — a behavioral evaluation harness developed by PBH Applied Systems. Scores measure real agent-adjacent task performance across structured output, tool dispatch, multi-turn state retention, and multi-step planning families — not perplexity or benchmark leaderboard proxies.

📌 This is the full-precision F16 baseline repository. The evaluated quantized variant is published separately: Q4_K_M. This card documents reference behavior under full precision. The quantized card shows how Q4_K_M compression impacts behavioral fidelity.

🌍 1M-Context Extended Window. Qwen2.5-14B-Instruct-1M supports up to 1,048,576 tokens of context — a differentiator for document-heavy workloads, long conversation histories, and multi-document analysis. Evaluation context was capped to 8,192 tokens (common Modal limit); full 1M capability is untested here but preserved in the model.

---

Try the Live AI Agent Demo & Compare Models

Two complementary interfaces, both free:

1. Quick Demo — Single Model Testing

Launch the PBH Applied Systems AI Agent Demo →

Free · 3 example queries per agent type · No account required · No document upload

Test this model (or others) against three agent workflow templates:

• Reasoning & Analysis — Chain-of-thought decomposition and structured problem-solving
  Example queries: Should a startup build on cloud LLMs or self-host quantized models? · Analyze the trade-offs between model quantization and inference latency. · What are the cost implications of running models with extended context windows in production?

• Document Intelligence — Long-context information extraction and Q&A
  Example queries: Extract key clauses from a contract and summarize legal risks. · Analyze a research paper and identify the novel contributions. · Compare market analysis reports and highlight strategic differences.

• Code & Automation — API scaffolding, data transformation, and task completion
  Example queries: Build a REST API endpoint for user authentication and include rate limiting. · Transform CSV data into a production-ready database schema. · Write a Python script to validate and clean messy customer records.

Every query runs on private GPU infrastructure — your prompts, reasoning traces, and outputs never leave the system and are never sent to Frontier model providers or cloud vendors. This matters for organizations with sensitive data, compliance requirements, or internal knowledge that cannot leave the building.

2. Model Comparison Arena — Side-by-Side Evaluation

Launch the quant_eval Agent Arena Space →

Compare two models in real-time across all three agent types.

The Agent Arena lets you:

• Select any two evaluated models (F16 or quantized variants) and test them side-by-side on the same queries
• View execution traces for both agents, showing chain-of-thought reasoning and tool dispatch
• Check the Model Leaderboard — the Q4_K_M pass rate in each of the eight quant_eval capability families for every model with a published run, marked ▼ degraded or ▲ improved where McNemar's paired test against the model's own full-weight baseline is significant, ordered by how many families degraded
• Read the Methodology tab — explanation of quant_eval, the 8 test families, and how to request a full evaluation report for models not yet in the series

Why compare in the Arena:

The demo shows what this model can do. The Arena shows how this model compares to others. If you're deciding between F16 and Q4_K_M, or between 14B and other sizes in the series, run the same query in the Arena with both selected. You'll see the exact differences in reasoning quality, tool dispatch, and output coherence — not benchmark scores, but real agent behavior.

Evaluation-backed leaderboard: Every score in the Arena comes from published quant_eval results: the quant_eval Public Corpus (DOI 10.5281/zenodo.22009419) and, for showcase models such as Ministral-3-14B-Instruct-2512, sealed run bundles published as summaries. The numbers are not opinionated; they're measured.

---

Model Description

This repository contains the full-precision F16 GGUF of Qwen/Qwen2.5-14B-Instruct-1M, a 14-billion parameter instruction-tuned model developed by Alibaba Qwen Team (January 2025). This variant features an extended 1,048,576-token (1M) context window and multilingual instruction-following capability — a significant differentiator from the standard 32K context.

The F16 format preserves all original float16 weights without quantization. In the PBH Applied Systems evaluation pipeline, the F16 variant serves as the reference baseline: it is evaluated first, then the quantized variant is evaluated against the same fixture set, enabling direct measurement of quantization impact on behavioral fidelity.

For production deployments emphasizing multiple-choice extraction fidelity, F16 is the right choice (0.970 vs Q4_K_M's 0.820). For footprint and cost optimization, the quantized variant (Q4_K_M) offers substantial gains.

Key Characteristics

• Parameters: 14B
• Format: GGUF F16 (full precision)
• File size: 29.5 GB
• SHA256: de08ea9c41234ef83b7aacf07f9ebc3cbaa20ca8aeb5f6417758a8798660aaa9
• Context window: 1,048,576 tokens (1M extended)
• Evaluation context: 8,192 tokens (common Modal cap)
• Minimum VRAM (GPU inference): ~31 GB
• Recommended GPU tier: A100 40GB (eval hardware) · RTX 6000 Ada (48 GB) · 2× A10G
• Inference speed (eval hardware): avg 33.06 tokens/sec on Modal A100-40GB GPU (NVIDIA)
• Multilingual: English, Chinese, and other languages

---

PBH Applied Systems Evaluation — quant_eval v7.22.22

Evaluation conducted by PBH Applied Systems, LLC using quant_eval v7.22.22 Run ID: Qwen2.5_14B_Instruct_1M_20260815_220633 · Fixtures: v7.22.22 (SHA256: 0673b5e5...) Evaluation date: August 15, 2026 · Seed: unsupported (Modal); top_k=40, top_p=0.95 · Hardware: Modal GPU cluster · Published: August 15, 2026

Evaluation Data Published to Zenodo

All evaluation artifacts are published under the quant_eval Public Corpus (DOI: 10.5281/zenodo.22009419):

• Run Provenance (D4): 10.5281/zenodo.22010462 — Selected run manifest and rollup fields, artifact SHA-256 hashes, runner and decoding configurations
• Golden Oracle Fixtures (D3): 10.5281/zenodo.22010278 — Test case specifications and fixture version crosswalk
• Family Pass Rates (D6): 10.5281/zenodo.22010623 — Aggregate pass rates per family, both runners, 95% Wilson confidence intervals
• Per-Case Behavioral Results (D1): 10.5281/zenodo.22009799 — 19,200 per-case rows, both precisions, all signals, case-level pass/fail
• Throughput Telemetry (D2): 10.5281/zenodo.22009987 — Per-call generation metrics, wall-time, token counts
• Efficiency & Footprint (D7): 10.5281/zenodo.22010723 — Stored artifact size, compression ratio, tokens/sec, observed wall-time ratios with hardware scope

Evaluation Contract: canonical_agentic_contract_v7.22.18:evaluation

---

Comparability & Fixture Methodology

Fixtures Enable Valid Variant Comparison

The quant_eval evaluation harness uses a consistent fixture set to measure both full-precision and quantized variants. The F16 and Q4_K_M variants were evaluated under the same fixture set, enabling direct comparison.

Fixture Generation:

What this means: Results are directly comparable. Both runners used identical fixtures, enabling precise measurement of Q4_K_M behavioral change relative to F16.

Note on cross-run comparability: Fixture content is identical across all quant_eval runs in the series; only the version label differs (this run: v7.22.22; 7B run: v7.22.19). The 14B is directly comparable with the 32B run (same substrate, decoding, and contract). It is not directly comparable with the 7B, which ran on a local RTX 4090 under different decoding settings (top_p 1.0, seed applied, 4,096-token context) and contract v7.22.17.

---

Per-Family Pass Rates — Full-Weight Baseline

All rates include 95% Wilson score confidence intervals. N = 200 cases per family.

Source: quant_eval Family Pass Rates, DOI 10.5281/zenodo.22010623

---

Key Findings

Finding 1: Tool Dispatch and MCQ — 95% and 97%

Qwen2.5-14B at full precision performs strongly on tool-adjacent tasks. End-to-end toolcall reaches 0.950 (all 10 failures are tool-selection errors), and MCQ extraction reaches 0.970. Schema-only toolcall_only is lower at 0.880; all 24 failures are tool-name selection — argument construction passes on all 200 cases.

Finding 2: Stateful Conversation — Perfect on F16

Two-turn state tracking reaches 100% pass rate on F16. This model exhibits strong memory of context changes across turns, making it suitable for stateful multi-turn agentic workflows without additional scaffolding.

Finding 3: Hybrid Responses — Payload Semantics Are the Weak Point

Mixed_brief_json reaches 0.780 (44 of 200 cases fail). JSON parsing and schema validity hold on all 200 cases; the failures are in payload content — all 44 fail semantic_payload_ok, and 4 of those also fail answer_line_ok. The model produces well-formed JSON but the payload values do not always match the expected semantics.

---

Quantization Variants Comparison

See how this model performs across the quantized variant. All evaluated under the same fixture set.

*Throughput measured on Modal GPU cluster. Different hardware tiers used during evaluation (F16 on A100, Q4_K_M on A10G) introduce infrastructure-dependent variance. Speed ratio is not a pure quantization metric. On equivalent hardware, Q4_K_M throughput would be comparable or faster due to reduced memory bandwidth requirements.

See the full Q4_K_M card:

• Q4_K_M — 69.6% smaller, stateful parity, end-to-end toolcall improves (+4.0 pp, p = 0.008), but MCQ degrades significantly (−15.0 pp)

---

Artifact Provenance

The artifact was produced from Qwen/Qwen2.5-14B-Instruct-1M using a custom-built llama.cpp conversion pipeline developed by PBH Applied Systems, without modification to model weights.

Artifact verification: SHA256 hashes are recorded in the run manifest (DOI 10.5281/zenodo.22010462). Download the F16 GGUF and verify:

sha256sum qwen-2.5-14B-instruct-1m-gguf-F16.gguf
# Should output: de08ea9c41234ef83b7aacf07f9ebc3cbaa20ca8aeb5f6417758a8798660aaa9

---

Evaluation Methodology

quant_eval v7.22.22 is a behavioral evaluation harness developed by PBH Applied Systems. The two-run architecture evaluates the full-precision (F16) model first, then evaluates the quantized variant against the same fixture set, enabling direct measurement of behavioral degradation or improvement.

Fixture set: v7.22.22 evaluation split

• Total unique cases: 1,600 (200 per family × 8 families)
• Per-family test families and pass signals:

Evaluation infrastructure: Modal GPU cluster (nvidia/cuda:12.4.1) Runner: full_weight_modal_llama_cpp (llama.cpp via Python binding, version 0.3.20) · GPU: Modal A100-40GB Decoding: temperature 0.7, top_k=40, top_p=0.95 (seed unsupported on Modal) Context: 8,192-token context window (common Modal cap; full 1M available in production) Reproducibility: Run manifest, fixtures, and per-case results published to Zenodo under the quant_eval Public Corpus

---

Deployment Recommendations

✅ Deploy F16 when:

• MCQ extraction is critical. F16 maintains 97% pass rate; Q4_K_M degrades to 82% (−15.0 pp, p = 6.9e−08).
• Stateful agentic workflows are required. Perfect 100% on turn-2 state tracking.
• You have enterprise GPU tier. A100 or dual A10G.

⚠️ Avoid F16 if:

• Hardware is constrained to <31 GB VRAM. Q4_K_M fits in ~10 GB, enabling deployment on A10G, L4, and other mid-tier GPUs.
• Cost efficiency is the priority. 69.6% footprint reduction; end-to-end tool dispatch actually improves (+4.0 pp, p = 0.008); state tracking remains perfect.
• Speed comparison requires caution. F16 and Q4_K_M were evaluated on different Modal GPU tiers (F16 on A100, Q4_K_M on A10G), so the observed 0.96× ratio reflects infrastructure variance, not pure quantization cost. On equivalent hardware, throughput difference would be minimal or favor Q4_K_M due to reduced memory bandwidth requirements.

---

About PBH Applied Systems

PBH Applied Systems, LLC is an Oklahoma City–based applied machine learning and AI systems company specializing in production-grade model evaluation, quantization pipelines, agentic AI infrastructure, and scalable AI-driven application development. The organization emphasizes engineering rigor, reproducibility, and real-world deployment constraints — particularly in environments where performance, cost efficiency, and reliability must be balanced against available hardware and budget.

Founder & Principal AI/ML Systems Architect — Patrick Hill

Patrick Hill is the Founder & Principal AI/ML Systems Architect of PBH Applied Systems with 10+ years of experience delivering advanced analytics, predictive modeling, and decision-support solutions across high-stakes operational environments. Patrick holds a Master of Science in Software Engineering with concentrations in Artificial Intelligence and Machine Learning and a B.S. in Business Finance.

Technical expertise spans: Python, SQL, Linux, Pandas, NumPy, scikit-learn, PyTorch, TensorFlow/Keras, HuggingFace Transformers, GGUF, llama.cpp, BitsAndBytes, PEFT, QLoRA, Flask APIs, Docker, CI/CD, Jupyter, Databricks, and quantization strategies.

Published Author: Patrick is the author of Applied Machine Learning: Concepts, Tools, and Case Studies — a 1,200+ page practitioner-oriented textbook adopted as required reading for CSC 373 – Machine Learning at the University of Advancing Technology.

Core Service Areas: LLM optimization & deployment · AI evaluation frameworks · Agentic AI infrastructure · Scalable AI application development · ML pipeline design & analytics · Model & agent cataloging.

---

🔬 About quant_eval & This Evaluation Series

quant_eval is a behavioral evaluation harness developed by PBH Applied Systems, LLC. It measures real agent-adjacent task performance across structured output, tool dispatch, multi-turn state retention, and multi-step planning — not perplexity or leaderboard proxies. Every model published under pbhappliedsystems has been independently evaluated using quant_eval before being recommended for any production role.

See it in action: Live AI Agent Demo →

---

📋 Interested in Deeper Engagement?

This model card documents behavioral evidence under quant_eval. If you're evaluating this model for production deployment, need guidance on quantization impact for your specific workload, or want to understand deployment tradeoffs under your infrastructure constraints, PBH Applied Systems offers several paths:

Async Intake Form — https://pbhappliedsystems.com/contact

Describe your use case, infrastructure, and evaluation needs. Responses are reviewed asynchronously.

Available Services:

• Evaluation Report — A written behavioral audit: per-family pass rates, F16 vs. quantized delta analysis, failure cluster diagnostics, deployment recommendation, and a quantization-level suggestion for your hardware and latency constraints. Evidence-standard numbers only. Typical engagements: $2,500–$5,000.

• Sample report — Read a complete sample report: a real evaluation report for Ministral-3-14B-Instruct-2512, its BF16 baseline against Q4_K_M, Q5_K_M and Q8_0, written exactly as a client receives it. <!-- pbh-sample-report -->

• Starter-Kit Recommender — Configured agent templates (reasoning, document intelligence, code automation) with proof metrics from your specific workload. Includes model selection, quantization strategy, and deployment architecture guidance tied to your infrastructure.

Submit your details via the form above, and we'll route you to the appropriate engagement.

---

💬 Community

Questions about this model, a model or quant you want measured next, or your own quantization results to compare:

• Discord — PBH Applied Systems: #model-requests, #your-evals, #deployment-help
• GitHub Discussions — quant_eval Public Corpus tools: model requests and polls on what to measure next
• This repo — the Community tab for anything specific to these files

<!-- pbh-community -->

---

License

This GGUF repository inherits the license of the base model: Apache 2.0 — Qwen/Qwen2.5-14B-Instruct-1M

The quant_eval evaluation harness, fuzz prompt builder, and scoring implementation are proprietary to PBH Applied Systems, LLC and are not included in this repository. The golden oracle fixture set used in this evaluation is published under CC BY 4.0 as part of the quant_eval Public Corpus (D3, DOI 10.5281/zenodo.22010278).

---

GGUF conversion and behavioral evaluation performed by PBH Applied Systems, LLC · quant_eval v7.22.22 · Run ID: Qwen2.5_14B_Instruct_1M_20260815_220633 · Published 2026-08-15