Model datasheet · Qwen3

Qwen3 14B

The sweet spot of the dense Qwen3 line: near-32B quality in a 16GB-friendly footprint, with the family's strong long-context showing (RULER 94.6 non-thinking).

Vendor: Alibaba
Architecture: Dense · 14.8B
Context: 32,768 tokens
License: Apache-2.0
Released: 2025-04-29
Vision: No

→ GGUF (llama.cpp / LM Studio / Ollama) → MLX (Apple Silicon)

§1 Characteristics by quantization band

Q2–Q3 — 2–3 bit

7GB weights · e.g. Q3_K_M + KV: 1.3GB at 8K · 5GB at 32K · 5GB at 128K

— Not yet rated · contributions welcome —

Q4–Q5 — 4–5 bit

9GB weights · e.g. Q4_K_M , Q5_K_M , mlx-4bit + KV: 1.3GB at 8K · 5GB at 32K · 5GB at 128K

Math & reasoning

7/10

data checked Aug 2026

quant-degradation-community — KLD/perplexity delta vs bf16 @ Q4_K_M · aggregated · 2026-08-30
Editorial derivation: small loss at 4-bit.

Coding

6/10

data checked Aug 2026

quant-degradation-community — KLD/perplexity delta vs bf16 @ Q4_K_M · aggregated · 2026-08-30
Editorial derivation: small loss at 4-bit.

Tool calling / agentic

6/10

data checked Aug 2026

quant-degradation-community — community consensus @ Q4_K_M · aggregated · 2026-08-30
Editorial derivation: small loss at 4-bit.

Instruction following

7/10

data checked Aug 2026

quant-degradation-community — community consensus @ Q4_K_M · aggregated · 2026-08-30
Editorial derivation: IF is not quant-sensitive.

Q6–Q8 — 6–8 bit

12GB weights · e.g. Q6_K , Q8_0 + KV: 1.3GB at 8K · 5GB at 32K · 5GB at 128K

— Not yet rated · contributions welcome —

FP16 — full precision (fp16/bf16)

30GB weights · e.g. bf16 + KV: 1.3GB at 8K · 5GB at 32K · 5GB at 128K

Math & reasoning

8/10

data checked Aug 2026

vendor-model-card — AIME'24 (thinking) = 79.3 @ bf16 · vendor-reported · 2026-08-30
MATH-500: 96.8 thinking. Non-thinking: 31.7.

Coding

7/10

data checked Aug 2026

vendor-model-card — LiveCodeBench v5 (thinking) = 63.5 @ bf16 · vendor-reported · 2026-08-30
Codeforces rating 1766. Absent from Aider polyglot.

Tool calling / agentic

7/10

data checked Aug 2026

vendor-model-card — BFCL v3 (thinking) = 70.4 @ bf16 · vendor-reported · 2026-08-30
Non-thinking: 61.5.

Instruction following

7/10

data checked Aug 2026

vendor-model-card — IFEval strict prompt (thinking) = 85.4 @ bf16 · vendor-reported · 2026-08-30
Arena-Hard: 91.7.

Factuality

5/10

data checked Aug 2026

vectara-hallucination — HHEM hallucination rate % (lower is better) = 5.4 @ bf16 · aggregated · 2026-08-30
Grounds well; family-wide open-domain knowledge gap applies.

Long context

7/10

data checked Aug 2026

vendor-model-card — RULER avg (non-thinking) = 94.6 @ bf16 · vendor-reported · 2026-08-30
90.1 thinking — unusually, non-thinking is better here. Native 32K; 128K via YaRN.

§2 Known issues & what fixes them

ISSUE-01 mild Q2–Q3 / Q4–Q5 / Q6–Q8 / FP16 · instruction-following

Endless repetition under greedy decoding — the model card warns against it explicitly; quantized builds are more susceptible.

Fix

Prompt technique [strong]

Card samplers: temp 0.6 / top_p 0.95 / top_k 20 (thinking) or 0.7 / 0.8 / 20 (non-thinking); presence_penalty 0–2 for stubborn loops.

Evidence · 2 sources · community-consensus

Qwen3-14B card (greedy-decoding warning) (vendor, 2026-08-30)
llama.cpp issue: Qwen3 GGUF repetition (github, 2026-08-30)

ISSUE-02 moderate Q2–Q3 / Q4–Q5 / Q6–Q8 / FP16 · factuality

Family-wide math/code tilt at the expense of world knowledge — grounding is good, recall is not.

Fix

RAG (retrieval) [strong]

Its 94.6% HHEM consistency means retrieval fixes what recall lacks.

Evidence · 2 sources · community-consensus

Family-wide knowledge-gap discussion (other, 2026-08-30)
Qwen3 release analysis (SimpleQA weakness) (blog, 2026-08-30)