RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can I run Qwen AgentWorld 35B A3B locally?

Qwen AgentWorld 35B A3B by Qwen has 35B parameters (3.0B active per token, 256 experts, 8 used per token). At 4-bit with an 8K context it needs about 21.6 GB of GPU memory. It fits on 48 of 65 GPUs and Macs we track. See all models that fit in 24 GB of VRAM.

21.6 GBQ4_K_M · 8K context
RTX 5090smallest consumer GPU that fits
M3 Pro 36GBsmallest Mac that fits
256Kmax context
GQAattention type

Memory needed by quant and context length

GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from unsloth/Qwen-AgentWorld-35B-A3B-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx32K ctx128K ctx
2-bit (Q2_K)rough13.1 GB13.1 GB13.6 GB15.5 GB
3-bit (Q3_K_M) •usable16.5 GB16.6 GB17.0 GB18.9 GB
4-bit (Q4_K_M) •good, most popular21.6 GB21.6 GB22.1 GB24.0 GB
5-bit (Q5_K_M) •very good25.6 GB25.7 GB26.1 GB28.0 GB
6-bit (Q6_K) •near lossless28.2 GB28.3 GB28.8 GB30.7 GB
8-bit (Q8_0) •effectively lossless35.3 GB35.4 GB35.9 GB37.7 GB
16-bit (F16/BF16) •original weights65.6 GB65.6 GB66.1 GB68.0 GB

Which GPUs and Macs run Qwen AgentWorld 35B A3B?

Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).

HardwareMemoryFits?Best quantNeedsMax tok/s (est.)
RTX 3060 12GB12 GBNo———details
RTX 4060 Ti 16GB16 GBFitsQ2_K13.1 GB256details
RTX 309024 GBFitsQ3_K_M16.6 GB648details
RTX 409024 GBFitsQ3_K_M16.6 GB698details
RTX 509032 GBFitsQ5_K_M25.7 GB783details
RTX 5070 Ti16 GBFitsQ2_K13.1 GB796details
RX 7900 XTX24 GBFitsQ3_K_M16.6 GB665details
Mac M4 Pro 48GB48 GBFitsQ6_K28.3 GB108details
Mac M4 Max 128GB128 GBFitsF1665.6 GB91details
Mac M3 Ultra 512GB512 GBFitsF1665.6 GB136details
DGX Spark128 GBFitsF1665.6 GB45details
RTX 407012 GBNo———details
RTX 5060 Ti 16GB16 GBFitsQ2_K13.1 GB398details
RTX 40608 GBNo———details
Arc B58012 GBNo———details
RTX 3060 Ti8 GBNo———details
RTX 30708 GBNo———details
RTX 3080 10GB10 GBNo———details
RTX 3080 12GB12 GBNo———details
RTX 3090 Ti24 GBFitsQ3_K_M16.6 GB698details
RTX 4060 Ti 8GB8 GBNo———details
RTX 4070 Super12 GBNo———details
RTX 4070 Ti12 GBNo———details
RTX 4070 Ti Super16 GBFitsQ2_K13.1 GB597details
RTX 408016 GBFitsQ2_K13.1 GB637details
RTX 4080 Super16 GBFitsQ2_K13.1 GB654details
RTX 50608 GBNo———details
RTX 507012 GBNo———details
RTX 508016 GBFitsQ2_K13.1 GB853details
2x RTX 30902×24 GBFitsQ8_035.4 GB293details
2x RTX 40902×24 GBFitsQ8_035.4 GB316details
2x RTX 50902×32 GBFitsQ8_035.4 GB561details
RTX A600048 GBFitsQ8_035.4 GB240details
RTX 6000 Ada48 GBFitsQ8_035.4 GB301details
RTX PRO 6000 Blackwell96 GBFitsF1665.6 GB298details
L40S48 GBFitsQ8_035.4 GB270details
A100 40GB40 GBFitsQ6_K28.3 GB613details
A100 80GB80 GBFitsF1665.6 GB340details
H100 PCIe80 GBFitsF1665.6 GB333details
H100 SXM80 GBFitsF1665.6 GB558details
H200141 GBFitsF1665.6 GB799details
RX 76008 GBNo———details
RX 7700 XT12 GBNo———details
RX 7800 XT16 GBFitsQ2_K13.1 GB555details
RX 7900 GRE16 GBFitsQ2_K13.1 GB512details
RX 7900 XT20 GBFitsQ3_K_M16.6 GB554details
RX 907016 GBFitsQ2_K13.1 GB569details
RX 9070 XT16 GBFitsQ2_K13.1 GB569details
Ryzen AI Max+ 395 128GB128 GBFitsF1665.6 GB43details
Arc A770 16GB16 GBFitsQ2_K13.1 GB498details
Mac M1 16GB16 GBNo———details
Mac M2 24GB24 GBFitsQ2_K13.1 GB89details
Mac M3 24GB24 GBFitsQ2_K13.1 GB89details
Mac M4 16GB16 GBNo———details
Mac M4 32GB32 GBFitsQ3_K_M16.6 GB83details
Mac M1 Pro 32GB32 GBFitsQ3_K_M16.6 GB139details
Mac M3 Pro 36GB36 GBFitsQ4_K_M21.6 GB78details
Mac M4 Pro 24GB24 GBFitsQ2_K13.1 GB243details
Mac M4 Pro 64GB64 GBFitsQ8_035.4 GB85details
Mac M1 Max 64GB64 GBFitsQ8_035.4 GB125details
Mac M2 Max 96GB96 GBFitsQ8_035.4 GB125details
Mac M3 Max 128GB128 GBFitsF1665.6 GB67details
Mac M4 Max 64GB64 GBFitsQ8_035.4 GB171details
Mac M2 Ultra 192GB192 GBFitsF1665.6 GB133details
Mac M3 Ultra 256GB256 GBFitsF1665.6 GB136details

Questions

How much VRAM does Qwen AgentWorld 35B A3B need?

About 21.6 GB at 4-bit (Q4_K_M) with an 8K context: 20.6 GB of weights (real GGUF file size), 0.2 GB of KV cache and 0.9 GB of runtime overhead. At 8-bit it needs about 35.4 GB.

What is the smallest GPU that runs Qwen AgentWorld 35B A3B?

At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RTX 5090 (32 GB).

Can I run Qwen AgentWorld 35B A3B on a Mac?

Yes. At Q4_K_M with 8K context it fits on a Mac M3 Pro 36GB or larger, using the default macOS GPU memory limit.

Similar models

Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.

Qwen3.6 35B A3B

36B · 22.1 GB at 4-bit

Qwen3.5 35B A3B

36B · 21.5 GB at 4-bit

Qwen3 32B

33B · 21.1 GB at 4-bit

Yi 34B 200K

34B · 21.6 GB at 4-bit

Yi 34B

34B · 21.6 GB at 4-bit

Yi 34B Chat

34B · 21.6 GB at 4-bit

Model facts

Hugging Face repoQwen/Qwen-AgentWorld-35B-A3B
Parameters34,660,610,688 (35B) · ~3.0B active (from the model name)
Layers / hidden size40 / 2048
AttentionGQA (hybrid: 30 linear/state layers); 16 heads, 2 KV heads, head dim 256
KV cache per token (FP16)20.0 KiB
Max context262,144 tokens
Licenseapache-2.0
Released on HFJun 22, 2026
Downloads (30d) / likes9,143 / 749

Sources: Hugging Face model API · config.json (Qwen/Qwen-AgentWorld-35B-A3B) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.