Qwen3.8 2.4T A95B — Hardware Requirements & GPU Compatibility
ChatFunctions
Qwen3.8-2.4T-A95B is Alibaba's Qwen-Max-class Mixture-of-Experts model with 2.4 trillion total parameters and about 95 billion active per token, built on the Qwen3.5 architectural foundation with a hybrid layout mixing gated linear-attention and gated full-attention layers. It is a post-trained, agent-focused model for coding, professional work, research, and long-horizon multi-step tasks, with adjustable reasoning depth via a reasoning-effort setting. It natively supports a 262K token context window, extensible to roughly 1 million tokens, and is released under a custom Qwen license rather than a fully open one. At 2.4 trillion parameters, this is server and multi-GPU territory even at 4-bit; nearly everyone will use a hosted endpoint instead.
Qwen3.8 2.4T A95B runs on a GPU cluster
Q4_K_M needs ~1.5 TB · about 3 nodes of 8× H100 SXM 80 GB
Other
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Qwen3.8 2.4T A95B
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve Qwen/Qwen3.8-2.4T-A95B --max-model-len 16384 --tensor-parallel-size 8 --pipeline-parallel-size <nodes>Size the nodes from the table above
Download: Qwen3.8-2.4T-A95B-Q4_K_M.gguf1467.71 GBunsloth
Other ways
each opens step-by-stepRun in cloud
Rent 3× 8-GPU H100 SXM nodes · Q4_K_M
- Estimated hourly price
- ~$64.56/hr
- est., whole cluster
- GPUs
- 24 GPUs
- 1.9 TB of VRAM
How Much VRAM Does Qwen3.8 2.4T A95B Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 673.4 GB | 674.6 GB | 2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss | Runs on 6× H200 SXM 141 GB | |
| 734.5 GB | 735.7 GB | 2-bit · i-quant extra small — Importance-weighted 2-bit, extra small | Runs on 6× H200 SXM 141 GB | |
| 1040.3 GB | 1041.5 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Runs on 7× B200 192 GB | |
| 1193.2 GB | 1194.4 GB | 3-bit medium — 3-bit medium quantization | Runs on 7× B200 192 GB | |
| 1468.4 GB | 1469.6 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Runs on 3 nodes of 8× H100 SXM | |
| 1743.6 GB | 1744.8 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Runs on 4 nodes of 8× H100 SXM | |
| 2018.8 GB | 2020 GB | 6-bit, very good — 6-bit quantization, very good quality | Runs on 4 nodes of 8× H100 SXM | |
| 2446.9 GB | 2448 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Runs on 5 nodes of 8× H100 SXM |
Selected Q4_K_M
4-bit medium quantization1468.8GBat 16K context
- Weights
- 1467.7 GB
- KV cache
- 0.8 GB
- Overhead
- 0.3 GB
Server-class: ~1.5 TB at 16K context · about 3 nodes of 8× H100 SXM 80 GB
Which Hardware Can Run Qwen3.8 2.4T A95B?
Q4_K_M · 1468.4 GBQwen3.8 2.4T A95B (Q4_K_M) requires 1468.4 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 1909+ GB is recommended. Using the full 262K context window can add up to 12.0 GB, bringing total usage to 1480.4 GB. No consumer GPU has enough memory.
Needs server-class hardware
Qwen3.8 2.4T A95B runs on a GPU cluster · Q4_K_M needs ~1.5 TB
| Hardware | Quant | Memory | Context | Hardware page | |
|---|---|---|---|---|---|
| 8× H200 SXM 141 GB · 1.1 TB | IQ3_XXS | 949 GB | 256K ctx | H200 SXM → | IQ3_XXS · 949 GB · 256K ctx |
| 8× B200 192 GB · 1.5 TB | IQ4_XS | 1315.9 GB | 256K ctx | B200 → | IQ4_XS · 1315.9 GB · 256K ctx |
| GPU cluster · 3 nodes of 8× H100 SXM 80 GB | Q4_K_M | 1468.8 GB | 256K ctx | H100 SXM → | Q4_K_M · 1468.8 GB · 256K ctx |
By VRAM tier · typical GPUs and Macs at Q4_K_M
Consumer GPUs (up to 32 GB) and Macs up to 512 GB: too small for this model (needs ≥ 673.8 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
Alternatives to Qwen3.8 2.4T A95B
Qwen 3.8 family · 12 models →Specifications
- Publisher
- Alibaba
- Family
- Qwen 3.8
- Parameters
- 2446.2B
- Active parameters
- 122B per token
- Architecture
- Qwen3_5MoeForCausalLM
- Context Length
- 262,144 tokens
- Vocabulary Size
- 248,320
- Release Date
- 2026-08-08
- License
- Other
Frequently Asked Questions
How much VRAM does Qwen3.8 2.4T A95B need?
Qwen3.8 2.4T A95B requires 1468.4 GB of VRAM at Q4_K_M, or 4893.1 GB at BF16. Full 262K context adds up to 12.0 GB (1480.4 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 2446.2B × 4.8 bits ÷ 8 = 1467.7 GB
KV Cache + Overhead ≈ 0.7 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 12.7 GB (at full 262K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run Qwen3.8 2.4T A95B?
No — Qwen3.8 2.4T A95B requires at least 673.4 GB at IQ2_XXS, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
What's the best quantization for Qwen3.8 2.4T A95B?
For Qwen3.8 2.4T A95B, Q4_K_M (1468.4 GB) offers the best balance of quality and VRAM usage. Q5_K_M (1743.6 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 673.4 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Qwen3.8 2.4T A95B on a Mac?
Qwen3.8 2.4T A95B requires at least 673.4 GB at IQ2_XXS. None of the 146 Macs we list has that much usable memory (the largest has 508.5 GB after macOS overhead).
Can I run Qwen3.8 2.4T A95B locally?
Yes — Qwen3.8 2.4T A95B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
What's the download size of Qwen3.8 2.4T A95B?
At Q4_K_M, the download is about 1467.71 GB. The full-precision BF16 version is 4892.37 GB. The smallest option (IQ2_XXS) is 672.70 GB.
Which GPUs can run Qwen3.8 2.4T A95B?
No single consumer GPU has enough VRAM to run Qwen3.8 2.4T A95B at Q4_K_M (1468.4 GB). Multi-GPU or professional hardware is required.
Which devices can run Qwen3.8 2.4T A95B?
Qwen3.8 2.4T A95B requires at least 1468.4 GB at Q4_K_M, which exceeds the usable memory of every unified-memory device we list (the largest has 508.5 GB).