How much VRAM does Qwen3.5 122B A10B need?
Qwen3.5 122B A10B needs 70.6 GB for its weights at 4-bit, 72.0 GB with an 8K context and 72.5 GB with 32K. At 8-bit, 125 GB.
- Parameters
- 125B
- Active per token
- 10B
- Longest context
- 256K
- Released
- Feb 2026
- License
- Apache 2.0
- Strongest at
- Chat, Coding, Agents, Docs
Memory at each precision
| Precision | Weights | With 8K | With 32K | With 128K |
|---|---|---|---|---|
| 16-bit | 233 GB | 234 GB | 235 GB | 237 GB |
| 8-bit | 124 GB | 125 GB | 126 GB | 128 GB |
| 6-bit | 95.5 GB | 96.9 GB | 97.4 GB | 99.7 GB |
| 5-bit | 82.9 GB | 84.2 GB | 84.8 GB | 87.0 GB |
| 4-bit | 70.6 GB | 72.0 GB | 72.5 GB | 74.8 GB |
| 3-bit | 56.9 GB | 58.3 GB | 58.8 GB | 61.1 GB |
| 2-bit | 43.7 GB | 45.0 GB | 45.6 GB | 47.8 GB |
| 1.6-bit | 24.0 GB | 25.4 GB | 25.9 GB | 28.2 GB |
One conversation, 16-bit context cache, runtime overhead included.
What runs it
| Machine | Best fit | Context | Speed | First reply |
|---|---|---|---|---|
| RTX 3060 12 GB | Too big | |||
| RTX 4060 Ti 16 GB | Too big | |||
| RTX 3090 | Too big | |||
| RTX 4090 | Too big | |||
| RTX 5090 | Too big | |||
| 2× RTX 3090 | Too big | |||
| RTX PRO 6000 Blackwell | 5-bit | 256K | 115 tok/s | 0.5 s |
| M4 Max 128 GB | 5-bit | 256K | 37 tok/s | 10 s |
| M5 Max 128 GB | 5-bit | 256K | 40 tok/s | 2.9 s |
| Ryzen AI Max+ 395 128 GB | 5-bit | 256K | 19 tok/s | 5.8 s |
| DGX Spark 128 GB | 6-bit | 256K | 21 tok/s | 2.1 s |
| M3 Ultra 256 GB | 8-bit | 256K | 37 tok/s | 5.6 s |
| M5 Ultra 512 GB | 8-bit | 256K | 49 tok/s | 1.8 s |
| H100 SXM | 4-bit | 176K | 181 tok/s | 0.4 s |
| DGX Station GB300 | 8-bit | 256K | 198 tok/s | 0.2 s |
| 8× H200 NVL | 8-bit | 256K | 164 tok/s | 0.1 s |
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 qwen3.5:122b-a10b- llama.cpp
llama-server -hf unsloth/Qwen3.5-122B-A10B-GGUF- vLLM
vllm serve Qwen/Qwen3.5-122B-A10B- MLX (Mac)
mlx_lm.chat --model mlx-community/Qwen3.5-122B-A10B-4bit
Questions
How much VRAM does Qwen3.5 122B A10B need?
Qwen3.5 122B A10B needs 70.6 GB for its weights at 4-bit, 72.0 GB with an 8K context and 72.5 GB with 32K. At 8-bit, 125 GB.
Can Qwen3.5 122B A10B run on an RTX 4090?
Not on one card. The smallest machine here that runs it is the RTX PRO 6000 Blackwell.