Models / Qwen3
Qwen3 by Qwen
14 models (0.6B to 235B) · ↓ 86.5M downloads a month · from $0.08/h on GTX 1660 SUPER right now. Each model's versions do the same job; quantized ones (AWQ, FP8, 4-bit) need less GPU memory and run on cheaper machines.
Qwen3 0.6B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-0.6B-Base | vLLM | 1.2 GB | 4.6 GB | $0.10/h RTX 3060 |
| InstructQwen/Qwen3-0.6B | vLLM | 1.5 GB | 4.9 GB | $0.10/h RTX 3060 |
| Qwen3Guard Gen 0.6BQwen/Qwen3Guard-Gen-0.6B | vLLM | 1.5 GB | 4.9 GB | $0.10/h RTX 3060 |
| FP8Qwen/Qwen3-0.6B-FP8 | vLLM | 1.1 GB | 4.4 GB | $0.10/h RTX 3060 |
| unsloth bnb 4bit (unsloth)unsloth/Qwen3-0.6B-unsloth-bnb-4bit | vLLM | 580 MB | 3.9 GB | $0.10/h RTX 3060 |
| ONNX (onnx-community)onnx-community/Qwen3-0.6B-ONNX | ONNX Runtime GenAI | 520 MB | 3 GB | $0.08/h GTX 1660 SUPER |
| quantized.w4a16 (RedHatAI)RedHatAI/Qwen3-0.6B-quantized.w4a16 | vLLM | 860 MB | 4.2 GB |
Qwen3 4B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-4B-Base | vLLM | 8.0 GB | 12.5 GB | $0.14/h RTX A4000 |
| InstructQwen/Qwen3-4B | vLLM | 8.0 GB | 12.5 GB | $0.14/h RTX A4000 |
| AWQQwen/Qwen3-4B-AWQ | vLLM | 2.7 GB | 6.5 GB | $0.10/h RTX 3060 |
| FP8Qwen/Qwen3-4B-FP8 | vLLM | 5.2 GB | 9.3 GB | $0.10/h RTX 3060 |
| Qwen3Guard Gen 4BQwen/Qwen3Guard-Gen-4B | vLLM | 8.8 GB | 13.4 GB | $0.14/h RTX A4000 |
| unsloth bnb 4bit (unsloth)unsloth/Qwen3-4B-unsloth-bnb-4bit | vLLM | 3.5 GB | 7.5 GB | $0.10/h RTX 3060 |
| Base unsloth bnb 4bit (unsloth)unsloth/Qwen3-4B-Base-unsloth-bnb-4bit | vLLM |
Qwen3 4B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-4B-Instruct-2507 | vLLM | 8.0 GB | 12.5 GB | $0.14/h RTX A4000 |
| FP8Qwen/Qwen3-4B-Instruct-2507-FP8 | vLLM | 5.2 GB | 9.3 GB | $0.10/h RTX 3060 |
| unsloth bnb 4bit (unsloth)unsloth/Qwen3-4B-Instruct-2507-unsloth-bnb-4bit | vLLM | 3.5 GB | 7.5 GB | $0.10/h RTX 3060 |
| Qwen3 Sex (ystemsrx)ystemsrx/Qwen3-Sex | vLLM | 8.0 GB | 12.5 GB | $0.14/h RTX A4000 |
| bnb 4bit (unsloth)unsloth/Qwen3-4B-Instruct-2507-bnb-4bit | vLLM | 2.6 GB | 6.5 GB | $0.10/h RTX 3060 |
| GPTQ Int8 (JunHowie)JunHowie/Qwen3-4B-Instruct-2507-GPTQ-Int8 | vLLM | 4.5 GB | 8.6 GB | $0.10/h RTX 3060 |
|
Qwen3 4B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-4B-Thinking-2507 | vLLM | 8.0 GB | 12.5 GB | $0.14/h RTX A4000 |
| FP8Qwen/Qwen3-4B-Thinking-2507-FP8 | vLLM | 5.2 GB | 9.3 GB | $0.10/h RTX 3060 |
Qwen3 8B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-8B-Base | vLLM | 16 GB | 21.8 GB | $0.18/h 2x RTX 3060 |
| InstructQwen/Qwen3-8B | vLLM | 16 GB | 21.8 GB | $0.18/h 2x RTX 3060 |
| AWQQwen/Qwen3-8B-AWQ | vLLM | 6.1 GB | 10.3 GB | $0.10/h RTX 3060 |
| FP8Qwen/Qwen3-8B-FP8 | vLLM | 9.4 GB | 14 GB | $0.14/h RTX A4000 |
| Qwen3Guard Gen 8BQwen/Qwen3Guard-Gen-8B | vLLM | 16 GB | 21.8 GB | $0.18/h 2x RTX 3060 |
| NVFP4 (nvidia)nvidia/Qwen3-8B-NVFP4 | vLLM | 6.4 GB | 10.7 GB | $0.10/h RTX 3060 |
| quantized.w4a16 (RedHatAI)RedHatAI/Qwen3-8B-quantized.w4a16 | vLLM | 6.1 GB | 10.3 GB |
Qwen3 14B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-14B-Base | vLLM | 30 GB | 36.5 GB | $0.45/h Nvidia A40 |
| InstructQwen/Qwen3-14B | vLLM | 30 GB | 36.5 GB | $0.45/h Nvidia A40 |
| AWQQwen/Qwen3-14B-AWQ | vLLM | 10.0 GB | 14.8 GB | $0.14/h RTX A4000 |
| FP8Qwen/Qwen3-14B-FP8 | vLLM | 16 GB | 21.9 GB | $0.18/h 2x RTX 3060 |
| unsloth bnb 4bit (unsloth)unsloth/Qwen3-14B-unsloth-bnb-4bit | vLLM | 11 GB | 16.1 GB | $0.18/h 2x RTX 3060 |
| NVFP4 (nvidia)nvidia/Qwen3-14B-NVFP4 | vLLM | 11 GB | 15.4 GB | $0.14/h RTX A4000 |
| FP8 (nvidia)nvidia/Qwen3-14B-FP8 | vLLM | 16 GB | 21.9 GB |
Qwen3 30B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-30B-A3B-Instruct-2507 | vLLM | 61 GB | 71 GB | $0.89/h 2x Nvidia A40 |
| FP8Qwen/Qwen3-30B-A3B-Instruct-2507-FP8 | vLLM | 31 GB | 37.8 GB | $0.45/h Nvidia A40 |
| AWQ (stelterlab)stelterlab/Qwen3-30B-A3B-Instruct-2507-AWQ | vLLM | 17 GB | 21.7 GB | $0.18/h 2x RTX 3060 |
| AWQ 4bit (cyankiwi)cyankiwi/Qwen3-30B-A3B-Instruct-2507-AWQ-4bit | vLLM | 18 GB | 23.2 GB | $0.24/h RTX 3090 |
| quantized.w4a16 (RedHatAI)RedHatAI/Qwen3-30B-A3B-Instruct-2507-quantized.w4a16 | vLLM | 17 GB | 21.6 GB | $0.18/h 2x RTX 3060 |
Qwen3 30B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-30B-A3B-Base | vLLM | 61 GB | 71 GB | $0.89/h 2x Nvidia A40 |
| InstructQwen/Qwen3-30B-A3B | vLLM | 61 GB | 71 GB | $0.89/h 2x Nvidia A40 |
| FP8Qwen/Qwen3-30B-A3B-FP8 | vLLM | 32 GB | 39.2 GB | $0.45/h Nvidia A40 |
| GPTQ Int4Qwen/Qwen3-30B-A3B-GPTQ-Int4 | vLLM | 17 GB | 21.9 GB | $0.24/h RTX 3090 |
| w8a8 (nytopop)nytopop/Qwen3-30B-A3B.w8a8 | vLLM | 31 GB | 37.8 GB | $0.45/h Nvidia A40 |
| NVFP4 (nvidia)nvidia/Qwen3-30B-A3B-NVFP4 | vLLM | 18 GB | 23.2 GB | $0.24/h RTX 3090 |
| AWQ (QuixiAI)QuixiAI/Qwen3-30B-A3B-AWQ | vLLM |
Qwen3 30B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-30B-A3B-Thinking-2507 | vLLM | 61 GB | 71 GB | $0.89/h 2x Nvidia A40 |
| FP8Qwen/Qwen3-30B-A3B-Thinking-2507-FP8 | vLLM | 31 GB | 37.8 GB | $0.45/h Nvidia A40 |
| AWQ 4bit (cyankiwi)cyankiwi/Qwen3-30B-A3B-Thinking-2507-AWQ-4bit | vLLM | 18 GB | 23.2 GB | $0.24/h RTX 3090 |
| AWQ (QuantTrio)QuantTrio/Qwen3-30B-A3B-Thinking-2507-AWQ | vLLM | 17 GB | 21.8 GB | $0.18/h 2x RTX 3060 |
Qwen3 32B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-32B | vLLM | 66 GB | 77.4 GB | $0.89/h 2x Nvidia A40 |
| AWQQwen/Qwen3-32B-AWQ | vLLM | 19 GB | 26.1 GB | $0.34/h 2x RTX 5060 Ti |
| FP8Qwen/Qwen3-32B-FP8 | vLLM | 34 GB | 42.7 GB | $0.45/h Nvidia A40 |
| NVFP4 (RedHatAI)RedHatAI/Qwen3-32B-NVFP4 | vLLM | 21 GB | 27.6 GB | $0.34/h 2x RTX 5060 Ti |
| unsloth bnb 4bit (unsloth)unsloth/Qwen3-32B-unsloth-bnb-4bit | vLLM | 39 GB | 48.3 GB | $0.78/h 4x RTX 4060 Ti |
| NVFP4 (nvidia)nvidia/Qwen3-32B-NVFP4 | vLLM | 21 GB | 27.6 GB | $0.34/h 2x RTX 5060 Ti |
| quantized.w4a16 (RedHatAI)RedHatAI/Qwen3-32B-quantized.w4a16 |
Qwen3 235B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-235B-A22B | vLLM | 470 GB | 526.4 GB | $11.77/h 8x RTX PRO 6000 |
| FP8Qwen/Qwen3-235B-A22B-FP8 | vLLM | 239 GB | 269.6 GB | $5.33/h 8x Nvidia A100 |
| abliterated FP8 (null-space)null-space/Qwen3-235B-A22B-abliterated-FP8 | vLLM | 236 GB | 266.7 GB | $5.33/h 8x Nvidia A100 |
Qwen3 235B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-235B-A22B-Instruct-2507 | vLLM | 470 GB | 526.4 GB | $11.77/h 8x RTX PRO 6000 |
| FP8Qwen/Qwen3-235B-A22B-Instruct-2507-FP8 | vLLM | 236 GB | 266.7 GB | $5.33/h 8x Nvidia A100 |
| AWQ (QuantTrio)QuantTrio/Qwen3-235B-A22B-Instruct-2507-AWQ | vLLM | 124 GB | 141.8 GB | $1.38/h 8x RTX 3090 |
Qwen3 235B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen3-235B-A22B-Thinking-2507 | vLLM | 470 GB | 526.4 GB | $11.77/h 8x RTX PRO 6000 |
| FP8Qwen/Qwen3-235B-A22B-Thinking-2507-FP8 | vLLM | 236 GB | 266.7 GB | $5.33/h 8x Nvidia A100 |
Model details come from Hugging Face and are refreshed daily; prices follow the live offers. GPU memory is GridPup's estimate: the weights, room for an 8,192-token conversation (chat models) and the runtime's own memory. Large models are split across the GPUs of one machine ("4x RTX 4090").