Skip to content

Qwen3 Coder Next 80B VRAM requirements

How much VRAM Qwen3 Coder Next 80B needs
4-bit weights: 45.0 GBWith 8K context: 46.0 GBWith 32K context: 46.6 GB8-bit, 8K context: 79.9 GB
Parameters
79.7B
Active per token
3B
Longest context
256K
Released
Jan 2026
License
Apache 2.0
Strongest at
Coding, Agents

How much VRAM Qwen3 Coder Next 80B needs

Memory Qwen3 Coder Next 80B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit148 GB149 GB150 GB152 GB
8-bit78.8 GB79.9 GB80.5 GB82.7 GB
6-bit60.8 GB61.9 GB62.5 GB64.7 GB
5-bit52.8 GB53.8 GB54.4 GB56.6 GB
4-bit45.0 GB46.0 GB46.6 GB48.9 GB
3-bit36.3 GB37.3 GB37.9 GB40.1 GB
2-bit27.8 GB28.9 GB29.4 GB31.7 GB
1.6-bit15.3 GB16.4 GB16.9 GB19.2 GB

One conversation, 16-bit context cache, runtime overhead included.

GPUs and Macs that run Qwen3 Coder Next 80B

The best way Qwen3 Coder Next 80B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Qwen3 Coder Next 80B on RTX 4060 report
RTX 5060Too bigPDF: Qwen3 Coder Next 80B on RTX 5060 report
RTX 3070Too bigPDF: Qwen3 Coder Next 80B on RTX 3070 report
RTX 3080 10 GBToo bigPDF: Qwen3 Coder Next 80B on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: Qwen3 Coder Next 80B on RTX 3060 12 GB report
Arc B580Too bigPDF: Qwen3 Coder Next 80B on Arc B580 report
RTX 4070Too bigPDF: Qwen3 Coder Next 80B on RTX 4070 report
RTX 4070 SuperToo bigPDF: Qwen3 Coder Next 80B on RTX 4070 Super report
RTX 5070Too bigPDF: Qwen3 Coder Next 80B on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: Qwen3 Coder Next 80B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: Qwen3 Coder Next 80B on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: Qwen3 Coder Next 80B on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: Qwen3 Coder Next 80B on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: Qwen3 Coder Next 80B on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: Qwen3 Coder Next 80B on RTX 4080 Super report
RTX 5070 TiToo bigPDF: Qwen3 Coder Next 80B on RTX 5070 Ti report
RTX 5080Too bigPDF: Qwen3 Coder Next 80B on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: Qwen3 Coder Next 80B on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: Qwen3 Coder Next 80B on RTX 3090 report
RTX 4090Too bigPDF: Qwen3 Coder Next 80B on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: Qwen3 Coder Next 80B on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: Qwen3 Coder Next 80B on RTX 5090 report
2× RTX 30903-bit256K191 tok/s0.3 sPDF: Qwen3 Coder Next 80B on 2× RTX 3090 report
2× RTX 40903-bit256K197 tok/s0.1 sPDF: Qwen3 Coder Next 80B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB3-bit256K221 tok/s0.3 sPDF: Qwen3 Coder Next 80B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada3-bit256K193 tok/s0.2 sPDF: Qwen3 Coder Next 80B on RTX 6000 Ada report
L40S3-bit256K184 tok/s0.3 sPDF: Qwen3 Coder Next 80B on L40S report
2× RTX 50905-bit256K215 tok/s0.1 sPDF: Qwen3 Coder Next 80B on 2× RTX 5090 report
4× RTX 30908-bit256K127 tok/s0.2 sPDF: Qwen3 Coder Next 80B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit256K182 tok/s0.3 sPDF: Qwen3 Coder Next 80B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: Qwen3 Coder Next 80B on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: Qwen3 Coder Next 80B on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: Qwen3 Coder Next 80B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: Qwen3 Coder Next 80B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: Qwen3 Coder Next 80B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB3-bit256K73 tok/s5.9 sPDF: Qwen3 Coder Next 80B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit256K68 tok/s2.9 sPDF: Qwen3 Coder Next 80B on M4 Max 128 GB report
M5 Max 128 GB8-bit256K72 tok/s0.9 sPDF: Qwen3 Coder Next 80B on M5 Max 128 GB report
M3 Ultra 96 GB6-bit256K87 tok/s1.7 sPDF: Qwen3 Coder Next 80B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit256K78 tok/s1.7 sPDF: Qwen3 Coder Next 80B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit256K78 tok/s1.7 sPDF: Qwen3 Coder Next 80B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit256K92 tok/s0.5 sPDF: Qwen3 Coder Next 80B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB3-bit256K70 tok/s1.7 sPDF: Qwen3 Coder Next 80B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit256K39 tok/s2.2 sPDF: Qwen3 Coder Next 80B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit256K54 tok/s1.6 sPDF: Qwen3 Coder Next 80B on RTX Spark report
DGX Spark8-bit256K50 tok/s1.8 sPDF: Qwen3 Coder Next 80B on DGX Spark report
Dual DGX Spark8-bit256K50 tok/s1.0 sPDF: Qwen3 Coder Next 80B on Dual DGX Spark report
DGX Station GB3008-bit256K283 tok/s0.1 sPDF: Qwen3 Coder Next 80B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit256K182 tok/s0.2 sPDF: Qwen3 Coder Next 80B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit256K182 tok/s0.1 sPDF: Qwen3 Coder Next 80B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit256K182 tok/s0.1 sPDF: Qwen3 Coder Next 80B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM6-bit256K215 tok/s0.4 sPDF: Qwen3 Coder Next 80B on A100 80 GB SXM report
H100 SXM6-bit256K252 tok/s0.1 sPDF: Qwen3 Coder Next 80B on H100 SXM report
4× H200 NVL8-bit256K260 tok/s0.1 sPDF: Qwen3 Coder Next 80B on 4× H200 NVL report
8× H200 NVL8-bit256K260 tok/s0.1 sPDF: Qwen3 Coder Next 80B on 8× H200 NVL report

Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.

How to run it

Ollama
ollama run qwen3-coder-next:latest
llama.cpp
llama-server -hf unsloth/Qwen3-Coder-Next-GGUF
vLLM
vllm serve Qwen/Qwen3-Coder-Next
MLX (Mac)
mlx_lm.chat --model mlx-community/Qwen3-Coder-Next-4bit

Questions

How much VRAM does Qwen3 Coder Next 80B need?

Qwen3 Coder Next 80B needs 45.0 GB for its weights at 4-bit, 46.0 GB with an 8K context and 46.6 GB with 32K. At 8-bit, 79.9 GB.

What is the minimum VRAM for Qwen3 Coder Next 80B?

About 37.3 GB at 3-bit with an 8K context, the lowest precision we recommend for a model this size. The smallest machine here that runs it without spilling is the 2× RTX 3090.

How big is Qwen3 Coder Next 80B?

79.7B parameters (3B active per token): 148 GB of weights at 16-bit, 45.0 GB at 4-bit.

Can Qwen3 Coder Next 80B run on an RTX 4090?

Not on one card. The smallest machine here that runs it is the 2× RTX 3090.

More models