model family

Qwen 3

Published by Qwen. 13 members have been measured on real machines. Every number below was signed by the machine that produced it — none of it is from a model card.

what it takes, what it returns

text → text

Declared by the network’s catalog, not measured — a signature says what this family would be asked. Whether it answered is the signed capability flags on each member.

hf.co/mradermacher/Qwen3-VL-8B-Instruct-abliterated-GGUF:Q2_K3 machines

4,096recalled

—served

22.4tok/s

hf.co/unsloth/Qwen3-VL-4B-Instruct-GGUF:Q4_03 machines · 4.02B · Q4_0

4,096recalled

—served

40.9tok/s

hf.co/unsloth/Qwen3-VL-4B-Instruct-GGUF:Q4_12 machines

4,096recalled

—served

37.0tok/s

hf.co/MaziyarPanahi/Qwen3-0.6B-GGUF:Q4_K_M1 machine · 752M · Q4_K_M

4,096recalled

—served

37.4tok/s

hf.co/unsloth/Qwen3-4B-GGUF:IQ3_XXS1 machine

4,096recalled

—served

0.9tok/s

hf.co/unsloth/Qwen3-VL-4B-Instruct-GGUF:IQ4_XS1 machine · 4.02B · IQ4_XS

4,096recalled

—served

6.4tok/s

hf.co/MaziyarPanahi/Qwen3-0.6B-GGUF:Q2_K1 machine

—recalled

—served

16.0tok/s

hf.co/MaziyarPanahi/Qwen3-0.6B-GGUF:Q3_K_M1 machine · 752M · Q3_K_M

—recalled

—served

44.6tok/s

hf.co/MaziyarPanahi/Qwen3-0.6B-GGUF:Q5_K_M1 machine

—recalled

—served

23.9tok/s

hf.co/unsloth/Qwen3-4B-GGUF:IQ1_S1 machine

—recalled

—served

24.9tok/s

hf.co/unsloth/Qwen3-4B-GGUF:Q4_11 machine

—recalled

—served

19.4tok/s

hf.co/unsloth/Qwen3-VL-4B-Instruct-GGUF:Q2_K1 machine

Builds this network knows of

Declared, not measured — this is what the catalog says exists and what it would cost a machine to hold. Nothing here is signed, and none of it claims the build works.

Qwen 3 30B-A3B

hf.co/unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF:IQ1_SIQ1_Sollama8.3 GiB on diskquantized by unsloth

hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q2_K_LQ2_K_Lollama10.6 GiB on diskquantized by unsloth

hf.co/unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF:Q3_K_MQ3_K_Mollama13.7 GiB on diskquantized by unsloth

hf.co/unsloth/Qwen3-30B-A3B-Thinking-2507-GGUF:Q4_K_XLQ4_K_XLollama16.5 GiB on diskquantized by unsloth

hf.co/unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF:Q5_K_XLQ5_K_XLollama20.2 GiB on diskquantized by unsloth

hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q8_K_XLQ8_K_XLollama33.5 GiB on diskquantized by unsloth

Qwen 3 4B

hf.co/unsloth/Qwen3-4B-GGUF:IQ1_SIQ1_Sollama1.0 GiB on diskquantized by unsloth

hf.co/unsloth/Qwen3-4B-GGUF:IQ3_XXSIQ3_XXSollama1.6 GiB on diskquantized by unsloth

hf.co/unsloth/Qwen3-4B-GGUF:Q4_0Q4_0ollama2.2 GiB on diskquantized by unsloth

hf.co/unsloth/Qwen3-VL-4B-Instruct-GGUF:Q4_1Q4_1ollama2.4 GiB on diskquantized by unsloth

hf.co/Qwen/Qwen3-4B-GGUF:Q6_KQ6_Kollama3.1 GiB on diskpublished by Qwen

hf.co/unsloth/Qwen3-VL-4B-Instruct-GGUF:BF16BF16ollama7.5 GiB on diskquantized by unsloth

Qwen 3 8B

hf.co/mradermacher/Qwen3-VL-8B-Instruct-abliterated-GGUF:Q2_KQ2_Kollama3.1 GiB on diskderived by mradermacher

hf.co/mradermacher/Qwen3-VL-8B-Instruct-abliterated-GGUF:IQ4_XSIQ4_XSollama4.3 GiB on diskderived by mradermacher

hf.co/Qwen/Qwen3-8B-GGUF:Q4_K_MQ4_K_Mollama4.7 GiB on diskpublished by Qwen

hf.co/AnkitAI/Parable-Qwen3-8B-Claude-Fable-5-GGUF:Q5_K_MQ5_K_Mollama5.4 GiB on diskderived by AnkitAI

hf.co/Qwen/Qwen3-8B-GGUF:Q6_KQ6_Kollama6.3 GiB on diskpublished by Qwen

hf.co/mradermacher/Qwen3-VL-8B-Instruct-abliterated-GGUF:F16F16ollama15.3 GiB on diskderived by mradermacher

Qwen 3 TTS

hf.co/Serveurperso/Qwen3-TTS-GGUF:Q4_K_MQ4_K_Mollama1.1 GiB on diskderived by Serveurperso

hf.co/Serveurperso/Qwen3-TTS-GGUF:Q8_0Q8_0ollama1.9 GiB on diskderived by Serveurperso

hf.co/Serveurperso/Qwen3-TTS-GGUF:BF16BF16ollama3.6 GiB on diskderived by Serveurperso

hf.co/Serveurperso/Qwen3-TTS-GGUF:F32F32ollama7.2 GiB on diskderived by Serveurperso

ask your own gram

Your concierge can answer this from the network rather than from this page. The ask carries the operator id, so it resolves to a capability and an engine — not to a name somebody has to recognise.

What has this network measured about the Qwen 3 family (epn.family.qwen-3), and can my gram run any of it? See /inferences/qwen-3

open a gramx with this ask →epn.family.qwen-3