Can I run Qwen3.8 2.4T A95B locally?
Qwen3.8 2.4T A95B by Qwen has 2.45T parameters (95B active per token, 512 experts, 10 used per token). At 4-bit with an 8K context it needs about 1394 GB of GPU memory. It fits on 0 of 65 GPUs and Macs we track.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from unsloth/Qwen3.8-2.4T-A95B-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | rough | 856 GB | 856 GB | 858 GB | 867 GB |
| 3-bit (Q3_K_M) | usable | 1115 GB | 1115 GB | 1117 GB | 1126 GB |
| 4-bit (Q4_K_M) | good, most popular | 1394 GB | 1394 GB | 1396 GB | 1405 GB |
| 5-bit (Q5_K_M) | very good | 1624 GB | 1625 GB | 1627 GB | 1636 GB |
| 6-bit (Q6_K) | near lossless | 1869 GB | 1870 GB | 1872 GB | 1880 GB |
| 8-bit (Q8_0) • | effectively lossless | 2423 GB | 2423 GB | 2425 GB | 2434 GB |
| 16-bit (F16/BF16) • | original weights | 4558 GB | 4559 GB | 4561 GB | 4569 GB |
Which GPUs and Macs run Qwen3.8 2.4T A95B?
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).
Prefer an API?
Qwen: Qwen3.8 2.4T A95B is on OpenRouter at $2.00 per 1M input tokens and $6.00 per 1M output tokens. Price details and history →
Questions
How much VRAM does Qwen3.8 2.4T A95B need?
About 1394 GB at 4-bit (Q4_K_M) with an 8K context: 1393 GB of weights (estimated), 0.7 GB of KV cache and 0.9 GB of runtime overhead. At 8-bit it needs about 2423 GB.
What is the smallest GPU that runs Qwen3.8 2.4T A95B?
No single consumer GPU in our table fits it at Q4_K_M with 8K context.
Can I run Qwen3.8 2.4T A95B on a Mac?
Not at Q4_K_M with 8K context on any Mac in our table, even a 512 GB Mac Studio, using the default GPU memory limit.
Similar models
Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.
Qwen3.8 Flash Next
180B · 104 GB at 4-bitQwen3.6 35B A3B
36B · 22.1 GB at 4-bitQwen3.5 35B A3B
36B · 21.5 GB at 4-bitKimi K3
2.78T · 1584 GB at 4-bitKimi K3.1 Pro
2.78T · 1584 GB at 4-bitLongCat 2.5
1.78T · 1014 GB at 4-bitModel facts
| Hugging Face repo | Qwen/Qwen3.8-2.4T-A95B |
| Parameters | 2,446,182,725,504 (2.45T) · ~95B active (from the model name) |
| Layers / hidden size | 92 / 8192 |
| Attention | GQA (hybrid: 69 linear/state layers); 64 heads, 4 KV heads, head dim 256 |
| KV cache per token (FP16) | 92.0 KiB |
| Max context | 262,144 tokens |
| License | other |
| Released on HF | Aug 8, 2026 |
| Downloads (30d) / likes | 28,798 / 1,398 |
Sources: Hugging Face model API · config.json (Qwen/Qwen3.8-2.4T-A95B) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.