What is Qwen3-235B-A22B Thinking (2507) good for?

Multi-step reasoning with a long chain of thought, which Alibaba recommends for highly complex tasks. Apache 2.0 weights, on a multi-GPU node.

When is Qwen3-235B-A22B Thinking (2507) not a good fit?

Short, cost-sensitive requests: it runs in thinking mode only, so thinking tokens land on the output rate. Text only, and self-hosting takes a multi-GPU node.

What is the cheapest way to run Qwen3-235B-A22B Thinking (2507)?

Hosted, unless you push serious volume. Alibaba Cloud charges $0.23 in / $2.30 out per 1M tokens. The cheapest rental that fits is 8x RTX 3090 at $979 a month, which costs the same as about 2B tokens a month on that API.

Can I self-host Qwen3-235B-A22B Thinking (2507)?

Yes. The weights are Apache 2.0 licensed. At 4-bit it needs about 137 GB of GPU memory, which starts at roughly $979 a month on the cheapest rental that fits.

Open weights · Apache 2.0

Qwen3-235B-A22B Thinking (2507)

July 2025 update of Qwen3-235B-A22B in reasoning mode, a 235B MoE with 22B active.

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: $0.23 /1M tokens
Output: $2.30 /1M tokens

3 providers

Key Specifications

Context window: 262K tokens
Max output: 66K tokens
Parameters: 235B, 22B 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	Cost at 10M in + 2M out
Alibaba Cloud Creator Cheapest	$0.23	$2.30	$6.90	View
Novita	$0.30	$3.00	$9.00	View
Lyceum	$0.50	$2.00	$9.00	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

Qwen3-235B-A22B Thinking (2507) needs about 137 GB of GPU memory at 4-bit with 32K context. The cheapest rental that fits is 8× RTX 3090 at about $979 a month.

That costs the same as roughly 2B tokens a month on Alibaba Cloud's API. Self-hosting is more expensive below that volume.

Single node ≤ 8

Precision	Memory	Cheapest fit (32K context)(Full 262K context)	Cost /mo	Break-even vs API
4-bitINT4 / FP4	137 GB157 GB	8× RTX 3090	$979$2,419	2B tokens /mo4B tokens /mo
8-bitFP8 / INT8	243 GB288 GB	8× RTX A6000	$3,168$3,168	6B tokens /mo6B tokens /mo
16-bitFP16 / BF16	478 GB523 GB	8× A100	$10,138$10,138	18B tokens /mo18B tokens /mo

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 Qwen3-235B-A22B Thinking (2507) by blended rate, each at its own cheapest provider.

Model	Blended /1M	Input /1M	Output /1M	Context	Cutoff	vs Qwen3-235B-A22B Thinking (2507)
Qwen3-VL-235B-A22B Instruct Alibaba Cloud	$0.50	$0.30	$1.50	262K		−13%
DeepSeek V4 Pro DeepSeek	$0.5233	$0.45	$0.89	1M		−9%
GPT-5 Mini OpenAI	$0.5417	$0.25	$2.00	400K	May 2024	−6%
DeepSeek V3 Turbo DeepSeek	$0.55	$0.40	$1.30	64K		−4%
Qwen3-VL-30B-A3B Thinking Alibaba Cloud	$0.5667	$0.20	$2.40	262K		−1%
Qwen3-235B-A22B Thinking (2507) This model Alibaba Cloud	$0.575	$0.23	$2.30	262K
Qwen3-Coder-480B-A35B Alibaba Cloud	$0.575	$0.38	$1.55	262K		0%
GPT-4.1 Mini OpenAI	$0.60	$0.40	$1.60	1M	Jun 2024	+4%
Gemini 2.5 Flash Google Cloud	$0.6667	$0.30	$2.50	1M	Jan 2025	+16%
Gemini 3.5 Flash-Lite Google Cloud	$0.6667	$0.30	$2.50	1M	Mar 2026	+16%
GPT-3.5 Turbo OpenAI	$0.6667	$0.50	$1.50	16K	Sep 2021	+16%

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.

Lium

Built for agents: CLI, SDK and MCP server. Your agent rents and runs GPUs itself.

A permissionless marketplace. 1000+ nodes from independent providers worldwide.

Hard to find nodes available; B200 and B300 Blackwell nodes listed now.

Consumer + DC grade cards for cheap. H100s from $1/hr, RTX 5090s from $0.35/hr.

View Pricing

Frequently Asked Questions

What is Qwen3-235B-A22B Thinking (2507) good for?: Multi-step reasoning with a long chain of thought, which Alibaba recommends for highly complex tasks. Apache 2.0 weights, on a multi-GPU node.
When is Qwen3-235B-A22B Thinking (2507) not a good fit?: Short, cost-sensitive requests: it runs in thinking mode only, so thinking tokens land on the output rate. Text only, and self-hosting takes a multi-GPU node.
What is the cheapest way to run Qwen3-235B-A22B Thinking (2507)?: Hosted, unless you push serious volume. Alibaba Cloud charges $0.23 in / $2.30 out per 1M tokens. The cheapest rental that fits is 8x RTX 3090 at $979 a month, which costs the same as about 2B tokens a month on that API.
Can I self-host Qwen3-235B-A22B Thinking (2507)?: Yes. The weights are Apache 2.0 licensed. At 4-bit it needs about 137 GB of GPU memory, which starts at roughly $979 a month on the cheapest rental that fits.

More from Alibaba Cloud

Model	Context	Input /1M	Output /1M
Qwen3-Coder-480B-A35B Open weights	262K	$0.38	$1.55
Qwen3-VL-30B-A3B Thinking Open weights	262K	$0.20	$2.40
Qwen3-VL-235B-A22B Instruct Open weights	262K	$0.30	$1.50
Qwen3.6-Flash	1M	$0.25	$1.50
Qwen3.5-Plus	1M	$0.40	$2.40
Qwen3-VL-Plus	262K	$0.20	$1.60
Qwen3.7-Plus	1M	$0.28	$1.10
Qwen2.5-72B Instruct Open weights	33K	$0.38	$0.40