nvidia · qwen3_5_moe · based on Qwen/Qwen3.5-122B-A10B
Qwen3.5-122B-A10B-NVFP4
Qwen3.5-122B-A10B-NVFP4 by nvidia: a 64.6B-parameter mixture-of-experts open-weight model under the apache-2.0 license with tool calling and vision.
1,428,051 downloads · 54 likes · Model card · synced 2026-09-19
Specifications
- Parameters
- 64.6B (2B active)
- Architecture
- moe · qwen3_5_moe
- Layers
- 48
- Hidden size
- 3,072
- KV heads · head dim
- 2 · 256
- Vocabulary
- 248,320
- Native dtype
- —
- Context window
- 262,144 tokens
- Released
- 2026-05-13
- Last modified on Hub
- 2026-06-02
Features & licensing
- License
- apache-2.0
- Commercial use
- Yes
- Modalities
- text, vision
- Tool / function calling
- Yes
- Reasoning mode
- Yes
- Languages
- EN
- Quantised variants on Hub
- FP8, GGUF
- Pipeline
- text-generation
Task fit
Editorial scores (0–100) used by the recommendation engine.
- Chat70
- RAG / Q&A70
- Code70
- Summarisation70
- Extraction70
- Agents70
Hardware to run Qwen3.5-122B-A10B-NVFP4
Weights need about 142 GB at FP16, 75 GB at INT8 and 43 GB at INT4. The KV cache adds roughly 98 MB per 1,000 tokens per request. For a reference workload of 5 requests per second with 1,500 input and 300 output tokens, total GPU memory is around 58.2 GB, which fits on 1 × NVIDIA A100 (80 GB).
| Configuration | VRAM | Utilisation | Est. first token | Cloud / month |
|---|---|---|---|---|
| 3 × NVIDIA L4 (24 GB) | 72 GB | 81 % | ~0.5 s | $1,752 |
| 2 × NVIDIA L40S (48 GB) | 96 GB | 61 % | ~0.2 s | $2,774 |
| 1 × NVIDIA A100 (80 GB)recommended | 80 GB | 73 % | ~0.1 s | $2,336 |
| 1 × NVIDIA H100 (80 GB) | 80 GB | 73 % | ~0.0 s | $3,285 |
Estimates only. Run an assessment for your own traffic.
Auto-imported from Hugging Face. Architecture, license and feature data are synced from the model repository; the task-fit scores below are provisional defaults until a curator reviews them.