What is Qwen3.8-2.4T-A95B good for?

Self-hosted access to the checkpoint behind Qwen3.8-Max: a 2.4T-parameter mixture of experts with 95B active, thinking mode only, under the custom Qwen3.8-Max License.

When is Qwen3.8-2.4T-A95B not a good fit?

Image or video input: it is text only, and Alibaba's card says thinking cannot be turned off. Serving it takes more than one node, and the licence is Alibaba's own rather than Apache 2.0.

What is the cheapest way to run Qwen3.8-2.4T-A95B?

Hosted, unless you push serious volume. Alibaba Cloud charges $2.00 in / $6.00 out per 1M tokens. The cheapest rental that fits is 8x B300 at $45,331 a month, which costs the same as about 17B tokens a month on that API.

Can I self-host Qwen3.8-2.4T-A95B?

Yes. The weights are released under the Qwen3.8-Max License. At 4-bit it needs about 1,324 GB of GPU memory, which starts at roughly $45,331 a month on the cheapest rental that fits.

Open weights · Qwen3.8-Max License

Qwen3.8-2.4T-A95B

The open-weight checkpoint behind Qwen3.8-Max, a 2.4T-parameter MoE with 95B active over hybrid Gated DeltaNet and Gated Attention layers, under a custom Qwen3.8-Max licence. Thinking mode only.

Jump to: Specs Pricing Self-hosting Similar models FAQ More from Alibaba Cloud

Visit website Compare vs other LLMs →

List price via Alibaba Cloud

Input: $2.00 / 1M tokens
Output: $6.00 / 1M tokens

4 providers

Key Specifications

Context window: 262K tokens
Max output: 131K tokens
Released: Aug 2026
Parameters: 2.4T, 95B active
Inputs: Text

Capabilities Show details: Reasoning Thinks before it answers, always on or as a switchable mode. Function calling Connect to external tools, APIs, and systems. Structured output Return responses in structured formats like JSON.

Hosted API pricing

Input tokens M Output tokens M

Provider	Input / 1M tokens	Output / 1M tokens	Cached input / 1M	Cost at 10M in + 2M out
Alibaba Cloud Creator	$2.00	$6.00		$32.00	View
Novita	$2.00	$6.00	$0.25	$32.00	View
Together AI	$2.00	$6.00	$0.25	$32.00	View
Lyceum	$2.50	$6.00		$37.00	View

Heads up: Base-tier, on-demand rates per 1M tokens; cached, batch and long-context tiers excluded. A provider may serve a shorter context or a quantized build than the creator's release. Verify before provisioning. More on how we price.

Estimated cost to self-host

Qwen3.8-2.4T-A95B needs about 1,324 GB of GPU memory at 4-bit with 32K context. The cheapest rental that fits is 8× B300 at about $45,331 a month.

That costs the same as roughly 17B tokens a month on Alibaba Cloud's API. Self-hosting is more expensive below that volume.

Precision	Cheapest, 32K context	Cheapest, full 262K context
4-bitINT4 / FP4	8× B300 1,324 GB VRAM · $45,331/mo	8× B300 1,334 GB VRAM · $45,331/mo
8-bitFP8 / INT8	Needs 2,405 GB	Needs 2,427 GB
16-bitFP16 / BF16	Needs 4,805 GB	Needs 4,827 GB

Estimates based on median on-demand rates for Nvidia GPUs. Memory is weights plus KV cache for one request, using FP8 KV cache where supported and FP16/BF16 otherwise. Break-even assumes a 5:1 input-to-output ratio. No guarantee of runtime support, usable performance, or that a matching quantized build exists. How we estimate costs.

Similarly priced models

The models nearest Qwen3.8-2.4T-A95B by blended rate, each at its own cheapest provider.

Model	Blended / 1M	Input / 1M	Output / 1M	Context	Cutoff	vs Qwen3.8-2.4T-A95B
GLM-5.1 Z.AI	$1.80	$1.30	$4.30	205K		−33%
GLM-5.3 Z.AI	$1.90	$1.40	$4.40	1M		−29%
Mistral Medium 3.5 Mistral	$2.50	$1.50	$7.50	256K		−6%
Grok 4.5 xAI	$2.6667	$2.00	$6.00	500K	Feb 2026	0%
Grok 4.6 xAI	$2.6667	$2.00	$6.00	500K	Jan 2026	0%
Qwen3.8-2.4T-A95B This model Alibaba Cloud	$2.6667	$2.00	$6.00	262K
Qwen3.8-Max Alibaba Cloud	$2.6667	$2.00	$6.00	1M		0%
Gemini 2.5 Pro Google Cloud	$2.7083	$1.25	$10.00	1M	Jan 2025	+2%
GPT-5 OpenAI	$2.7083	$1.25	$10.00	400K	Sep 2024	+2%
GPT-5.1 OpenAI	$2.7083	$1.25	$10.00	400K	Sep 2024	+2%
Gemini 3.5 Flash Google Cloud	$2.75	$1.50	$9.00	1M		+3%

Prices are USD per 1M tokens at each model's cheapest listed provider. Blended is the cost of 10M input plus 2M output tokens, spread over the 12M.

Vast.ai

Our sponsor

Per-second billing. Pay only for what you use.

50+ GPU models from RTX 3060 to B200, from $0.04/hr

No minimums, no long-term contracts.

SOC2 certified · PyTorch, CUDA & TensorFlow templates ready

Learn more

Frequently Asked Questions

What is Qwen3.8-2.4T-A95B good for?: Self-hosted access to the checkpoint behind Qwen3.8-Max: a 2.4T-parameter mixture of experts with 95B active, thinking mode only, under the custom Qwen3.8-Max License.
When is Qwen3.8-2.4T-A95B not a good fit?: Image or video input: it is text only, and Alibaba's card says thinking cannot be turned off. Serving it takes more than one node, and the licence is Alibaba's own rather than Apache 2.0.
What is the cheapest way to run Qwen3.8-2.4T-A95B?: Hosted, unless you push serious volume. Alibaba Cloud charges $2.00 in / $6.00 out per 1M tokens. The cheapest rental that fits is 8x B300 at $45,331 a month, which costs the same as about 17B tokens a month on that API.
Can I self-host Qwen3.8-2.4T-A95B?: Yes. The weights are released under the Qwen3.8-Max License. At 4-bit it needs about 1,324 GB of GPU memory, which starts at roughly $45,331 a month on the cheapest rental that fits.

More from Alibaba Cloud

Model	Context	Input / 1M	Output / 1M
Qwen3.8-Max	1M	$2.00	$6.00
Qwen3-Coder-Plus	1M	$1.00	$5.00
Qwen3.7-Max	1M	$1.25	$3.75
Qwen3-Max	262K	$0.85	$3.38
Qwen3.5-397B-A17B Open weights	262K	$0.55	$3.50
Qwen3-VL-235B-A22B Thinking Open weights	262K	$0.40	$4.00
Qwen3.6-Plus	1M	$0.50	$3.00
Qwen3.5-122B-A10B Open weights	262K	$0.40	$3.20