96 GB VRAM (+16)
C
This setup is broadly balanced for this model.23.4 tok/s decode
~$9,999 MSRP
Can Qwen3.5 122B A10B run on NVIDIA A800 80GB?
YES — With Offload
C49Usable
○Estimated from fit model
Qwen3.5 122B A10B needs ~83.0 GB VRAM. NVIDIA A800 80GB has 80.0 GB. With Q3_K_M quantization, expect ~19 tok/s.
Runtime: llama.cppCapacity: OffloadBandwidth: HighStack: StandardBottleneck: Balanced
Share:
Operating mode
Choose the run profile you care about
Interactive favors responsiveness, while light API and scale-out lean harder on serving readiness. The fit stays the same, but the recommendation lens changes.
Current mode
Balanced
Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.
Select quantization to explore
F16 (Maximum quality) — 273.3 GB, exceeds 80.0 GB available
273.3 GB required80.0 GB available
342% VRAM needed
193.3 GB over capacity — needs offload or smaller quantization
Fit status
Too heavy
Decode
2.0 tok/s
TTFT
96800 ms
Safe context
4K
Memory
273.3 GB / 80.0 GB
Offload
70%
Memory breakdown
Weights250.1 GB
KV Cache14.3 GB
Runtime0.9 GB
Headroom8.0 GB
See how fast it feels
With memory offload — actual speed may be lowerSee how fast it feelsQwen3.5 122B A10B on NVIDIA A800 80GB
1st promptCold start — includes initialization
>What is local AI inference?
Local AI inference means running an AI language model directly on your own hardware — your laptop, desktop, or server — instead of sending requests to a remote cloud API.
When you run inference locally the model weights are loaded into your GPU or unified memory. Each token you generate requires reading those weights from memory, so memory bandwidth is the main bottleneck for decode speed.
Key benefits of running locally:
- Full privacy: your prompts never leave your machine
- No per-token cost or rate limits
- Works offline once the model is downloaded
- Latency depends only on your hardware
2nd promptWarm — model ready, no init overhead
>How much VRAM do I need?
It depends on the model size and quantization level. A rough rule of thumb:
Model size Q4 (4-bit) Q8 (8-bit) FP16
7B params ~4.3 GB ~7.5 GB ~14 GB
13B params ~7.9 GB ~13.9 GB ~26 GB
70B params ~42.7 GB ~74.9 GB ~140 GB
Most people use 4-bit quantization (Q4_K_M) which gives 90-95% of full quality at a fraction of the memory. A 24 GB GPU can comfortably run most 7B-13B models.
Estimated: 2.0 tok/s decode · 96.8s TTFT (warm) · 5 tok/s prefill
What limits this setup
This setup is broadly balanced for this model.
Very little memory headroom
You can run the model, but there is not much room left for longer context, bigger batches, extra apps, or future model updates.
Best improvement path
Buy headroom, not only minimum fit
A slightly larger memory tier gives you safer context growth and makes the recommendation more future-proof.
Performance by workload
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Chat | C | Tight fit | 23.5 tok/s | 4497 ms | 13K |
| Coding | C | Runs with offload (needs ~2.1 GB host RAM) | 18.9 tok/s | 10267 ms | 13K |
| Agentic Coding | F | Too heavy | 14.5 tok/s | 19414 ms | 13K |
| Reasoning | C | Runs with offload (needs ~2.1 GB host RAM) | 18.9 tok/s | 12134 ms | 13K |
| RAG | F | Too heavy | 14.5 tok/s | 24267 ms | 13K |
Quantization options
How Qwen3.5 122B A10B (122B params) fits at each quantization level on NVIDIA A800 80GB (80.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 47.6 GB | Low | C48 |
Q3_K_SBest for your GPU | 3 | 59.8 GB | Low | C48 |
NVFP4 | 4 | 68.3 GB | Medium | F0 |
Q4_K_M | 4 | 74.4 GB | Medium | F0 |
Q5_K_M | 5 | 87.8 GB | High | F0 |
Q6_K | 6 | 100.0 GB | High | F0 |
Q8_0 | 8 | 130.5 GB | Very High | F0 |
F16 | 16 | 250.1 GB | Maximum | F0 |
Get started
Copy-paste commands to run Qwen3.5 122B A10B on your machine.
Run
docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
--hf-repo "unsloth/Qwen3.5-122B-A10B-GGUF" \
--hf-file "Qwen3.5-122B-A10B-GGUF-Q3_K_M.gguf" \
-c 4096 -ngl 99Upgrade options