Can Qwen AgentWorld 35B A3B run on RTX 4090 24GB?

YES — With Offload

S86Excellent

Estimated — low-sample bucket· few comparable runs

Qwen AgentWorld 35B A3B needs ~24.8 GB VRAM. RTX 4090 24GB has 24.0 GB. With Q4_K_M quantization, expect ~80 tok/s.

Runtime: llama.cppCapacity: OffloadBandwidth: HighStack: StandardBottleneck: Balanced

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.

Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 24.8 GB, 80.0 tok/s, Runs with offload (needs ~0.7 GB host RAM)

24.8 GB required24.0 GB available

103% VRAM needed

0.8 GB over capacity — needs offload or smaller quantization

Fit status

Runs with offload (needs ~0.7 GB host RAM)

Decode

80.0 tok/s

TTFT

2419 ms

Safe context

Memory

24.8 GB / 24.0 GB

Memory breakdown

Weights21.2 GB

KV Cache0.3 GB

Runtime0.9 GB

Headroom2.4 GB

See how fast it feels

See how fast it feelsQwen AgentWorld 35B A3B on RTX 4090 24GB
1st promptCold start — includes initialization
>Solve: If a train travels at 60 mph for 2.5 hours, how far does it go?
Let me work through this step by step.

Given information:
  - Speed = 60 mph (miles per hour)
  - Time  = 2.5 hours

Formula: Distance = Speed x Time

Calculation:
  Distance = 60 mph x 2.5 h
           = 150 miles

The train travels 150 miles.

Double-check: at 60 mph the train covers 60 miles each hour. In half an hour it covers 30 miles. So in 2.5 hours: 60 + 60 + 30 = 150 miles. Confirmed.
2nd promptWarm — model ready, no init overhead
>What if the train stops for 20 minutes halfway?
The stop doesn't change the total distance covered while moving, but it affects total trip time.

Breakdown:
  - First half: 1.25 hours at 60 mph = 75 miles
  - Stop: 20 minutes (no distance)
  - Second half: 1.25 hours at 60 mph = 75 miles

Total distance: 150 miles (same)
Total time: 2.5 h + 0.33 h = 2.83 hours
Average speed: 150 / 2.83 = 53 mph

The distance stays the same but average speed drops to 53 mph because of the stop.
Estimated: 80.0 tok/s decode · 2.4s TTFT (warm) · 200 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	S	Runs with offload (needs ~0.5 GB host RAM)	81.1 tok/s	1302 ms	4K
Coding	S	Runs with offload (needs ~0.7 GB host RAM)	80.0 tok/s	2419 ms	4K
Agentic Coding	S	Runs with offload (needs ~0.9 GB host RAM)	78.0 tok/s	3610 ms	4K
Reasoning	S	Runs with offload (needs ~0.7 GB host RAM)	80.0 tok/s	2858 ms	4K
RAG	S	Runs with offload (needs ~0.9 GB host RAM)	78.0 tok/s	4512 ms	4K

Inference speed

Qwen AgentWorld 35B A3B inference speed — tokens per second by GPU & Mac

Estimated decode speed (tokens/sec) for Qwen AgentWorld 35B A3B at Q4_K_M across popular GPUs and Apple Silicon, using the fastest local runtime per device. Fastest is RTX 5090 32GB at ~140 tok/s. Speed is memory-bandwidth bound, so cards that fit the whole model in VRAM run far faster than ones that offload to system RAM.

GPU / Mac	Memory	Quant	Speed (tok/s)	Fits?
RTX 5090 32GB	32 GB	Q4_K_M	140.4	Fits
RTX 4090 24GB	24 GB	Q4_K_M	80.0	Offloads
Mac Studio M3 Ultra 256GB	256 GB	Q4_K_M	77.5	Fits
RX 7900 XTX 24GB	24 GB	Q4_K_M	67.5	Offloads
Mac Studio M2 Ultra 128GB	128 GB	Q4_K_M	64.6	Fits
RTX 3090 24GB	24 GB	Q4_K_M	64.0	Offloads
Mac Studio M1 Ultra 128GB	128 GB	Q4_K_M	61.2	Fits
MacBook Pro M4 Max 128GB	128 GB	Q4_K_M	47.9	Fits
MacBook Pro M4 Max 64GB	64 GB	Q4_K_M	47.9	Fits
MacBook Pro M3 Max 64GB	64 GB	Q4_K_M	36.7	Fits
MacBook Pro M1 Max 64GB	64 GB	Q4_K_M	33.7	Fits
MacBook Pro M4 Pro 48GB	48 GB	Q4_K_M	32.2	Fits
RTX 4080 Super 16GB	16 GB	Q4_K_M	29.1	Too big
RTX 4070 12GB	12 GB	Q4_K_M	10.2	Too big
RTX 3060 12GB	12 GB	Q4_K_M	6.0	Too big
RTX 4060 8GB	8 GB	Q4_K_M	4.4	Too big

Estimates for single-stream decoding at Q4_K_M; real tokens/sec varies with prompt length, context, batch size, and runtime build. Prompt processing (prefill) is faster than the decode figures shown here.

Quantization options

How Qwen AgentWorld 35B A3B (34.70000076293945B params) fits at each quantization level on RTX 4090 24GB (24.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q1_0_G128	1.125	5.0 GB	Very Low	A79
Q2_0_G128	1.71	9.3 GB	Low	A81
Q2_K	2	13.5 GB	Low	A82
Q3_K_SBest for your GPU	3	17.0 GB	Low	A82
NVFP4	4	19.4 GB	Medium	F0
Q4_K_M	4	21.2 GB	Medium	F0
Q5_K_M	5	25.0 GB	High	F0
Q6_K	6	28.5 GB	High	F0
Q8_0	8	37.1 GB	Very High	F0
F16	16	71.1 GB	Maximum	F0

Get started

Copy-paste commands to run Qwen AgentWorld 35B A3B on your machine.

Run

docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
  --hf-repo "Qwen/Qwen-AgentWorld-35B-A3B" \
  --hf-file "Qwen-AgentWorld-35B-A3B-Q4_K_M.gguf" \
  -c 4096 -ngl 99

Your hardware

More models your RTX 4090 24GB can run

Model	Params	Grade	Decode
Qwen 3.6 35B A3B	35B	A	53.4 tok/s
Qwen 3.5 35B A3B	35B	A	71.1 tok/s
Agents-A1 35B A3B	35.1B	S	77.7 tok/s

Frequently asked questions

Can RTX 4090 24GB run Qwen AgentWorld 35B A3B?

Yes, RTX 4090 24GB can run Qwen AgentWorld 35B A3B with a S grade (Runs with offload (needs ~0.7 GB host RAM)). Expected decode speed: 80.0 tok/s.

How much VRAM does Qwen AgentWorld 35B A3B need?

Qwen AgentWorld 35B A3B (34.70000076293945B parameters) requires approximately 24.8 GB of memory with Q4_K_M quantization.

What is the best quantization for Qwen AgentWorld 35B A3B?

The recommended quantization for Qwen AgentWorld 35B A3B is Q4_K_M, which balances quality and memory efficiency.

What speed will Qwen AgentWorld 35B A3B run at on RTX 4090 24GB?

On RTX 4090 24GB, Qwen AgentWorld 35B A3B achieves approximately 80.0 tokens per second decode speed with a time-to-first-token of 2419ms using Q4_K_M quantization.

Can RTX 4090 24GB run Qwen AgentWorld 35B A3B for coding?

For coding workloads, Qwen AgentWorld 35B A3B on RTX 4090 24GB receives a S grade with 80.0 tok/s and 4K context.

What context window can Qwen AgentWorld 35B A3B use on RTX 4090 24GB?

On RTX 4090 24GB, Qwen AgentWorld 35B A3B can safely use up to 4K tokens of context. The model's official context limit is 262K, but available memory constrains the safe maximum.

What should I upgrade first if Qwen AgentWorld 35B A3B feels slow on RTX 4090 24GB?

Buy headroom, not only minimum fit. A slightly larger memory tier gives you safer context growth and makes the recommendation more future-proof.

See all results for RTX 4090 24GB See all hardware for Qwen AgentWorld 35B A3B

Embed this result▼

Paste this snippet into any page to show a live fit card.

<iframe src="https://willitrunai.com/embed/qwen-agentworld-35b-a3b-on-rtx-4090-24gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>

Preview: