Skip to content
Llama Builds

Alibaba Qwen · Released July 2025

Qwen3 30B-A3B (2507)

What it takes to run Qwen3 30B-A3B (2507) locally: memory by quantization, the GPUs that handle it, and how fast it runs.

Updated

Parameters
30.5B (3.3B active)
Architecture
Mixture of experts
Context
256K tokens
License
Apache 2.0 license
Memory at Q4_K_M
~21GB

What do you need to run Qwen3 30B-A3B (2507)?

Qwen3 30B-A3B (2507) needs about 21GB of VRAM at Q4_K_M with 8K context (19GB for the weights plus KV cache and overhead). The smallest GPU that runs it comfortably at Q3_K_M is a 20GB card like the RTX 4000 Ada. For a good balance of price and speed we recommend the RTX 3090 (~135 tokens/sec). Because it is a mixture-of-experts model with 3.3B active parameters, it generates text much faster than a dense model of the same size, and it tolerates partial CPU offload better.

See the full hardware guide for Qwen3 30B-A3B (2507)

What to buy

GPUs for Qwen3 30B-A3B (2507)

Cheapest that runs it

RTX 4000 Ada

20GB · $1,250 MSRP

Quant
Q3_K_M
Speed (est.)
~84 tok/s
Runs well
Check price at Amazon

Best value

RTX 3090

24GB · ~$1,050 used

Quant
Q4_K_M
Speed (est.)
~135 tok/s
Runs well
Check price at Amazon

Fastest

RTX 5090

32GB · ~$4,300

Quant
Q6_K
Speed (est.)
~175 tok/s
Runs well
Check price at Amazon

Speeds are estimates at the quantization shown with 8K context. We may earn a commission when you buy through these links. Learn more.

Memory

Qwen3 30B-A3B (2507) VRAM requirements by quantization

Lower-bit quantizations shrink the model at some cost to quality. Q4_K_M is the usual sweet spot. Totals include the KV cache for the context length shown plus runtime overhead.

Memory needed for Qwen3 30B-A3B (2507) at each quantization level
QuantizationQualityWeightsTotal @ 8K ctxTotal @ 32K ctx
Q2_KRough12 GB13 GB16 GB
Q3_K_MUsable15 GB17 GB19 GB
Q4_K_MRecommendedDefault19 GB21 GB23 GB
Q5_K_MVery good22 GB24 GB26 GB
Q6_KNear-lossless25 GB28 GB30 GB
Q8_0Practically lossless33 GB35 GB37 GB
BF16Full precision61 GB64 GB67 GB

KV cache note: 2 x 48 layers x 4 KV heads x 128 dims x 2 bytes = 98,304 bytes (96 KiB) per token: about 0.8GB at 8K and 25.8GB at the full 262K context.

Performance

How fast Qwen3 30B-A3B (2507) runs

Single-user generation speed on popular GPUs that fit the model entirely in memory, at the best quantization each can hold.

  1. RTX 5090

    Q6_K · 32GB

    ~175tok/s

  2. RTX PRO 6000

    Q8_0 · 96GB

    ~150tok/s

  3. RTX 4090

    Q4_K_M · 24GB

    ~150tok/s

  4. RTX 3090

    Q4_K_M · 24GB

    ~135tok/s

  5. RTX PRO 5000 72GB

    Q8_0 · 72GB

    ~125tok/s

  6. RX 7900 XTX

    Q4_K_M · 24GB

    ~110tok/s

  7. RTX 6000 Ada

    Q8_0 · 48GB

    ~99tok/s

  8. V100 32GB

    Q6_K · 32GB

    ~92tok/s

  9. M5 Ultra 512GB

    Q8_0 · 512GB

    ~80tok/s

  10. M5 Ultra 256GB

    Q8_0 · 256GB

    ~80tok/s

Estimated (method)

Measured

Qwen3 30B-A3B (2507) benchmarks

Overview

About Qwen3 30B-A3B (2507)

Qwen3 30B-A3B was the breakout local model of 2025: a 30.5B mixture-of-experts with only 3.3B active parameters, so it runs fast even with part of it offloaded to system RAM. This entry covers the July 2025 (2507) refresh, which split it into separate Instruct and Thinking checkpoints and raised native context to 262K. It has been superseded by Qwen3.5 35B-A3B and then Qwen3.6 35B-A3B, which add vision and stronger agentic coding at the same speed. At Q4_K_M it needs about 21GB with 8K context, so it fits a 24GB card, and a 12GB or 16GB GPU with 32GB or more of system RAM still runs it at usable speed.

Strengths

  • +Very fast generation: 3.3B active parameters per token
  • +Tolerates CPU offload well on 12GB to 16GB GPUs
  • +262K native context with a modest 96 KiB per-token KV cache
  • +Separate Instruct and Thinking checkpoints, both Apache 2.0

Weaknesses

  • −Superseded by Qwen3.6 35B-A3B for coding and agents
  • −Less world knowledge than dense models of the same total size
  • −Text only
ChatReasoningCodingAgents & toolsMultilingualLong contextToggleable reasoning

Setup

How to run Qwen3 30B-A3B (2507)

Ollama
$ ollama run qwen3:30b-a3b-instruct-2507-q4_K_M
llama.cpp
$ llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_M -c 8192 -ngl 99

Weights: Qwen/Qwen3-30B-A3B-Instruct-2507 · GGUF: unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF

Alternatives

Models like Qwen3 30B-A3B (2507)

Alibaba Qwen

Qwen3.6 35B-A3B

35B (3B active) · 256K ctx

~24GBat Q4_K_M
CodingAgents & toolsReasoning

Google

Gemma 4 26B A4B

26B (3.8B active) · 256K ctx

~17GBat Q4_K_M
ChatVisionReasoning

Alibaba Qwen

Qwen3 Coder 30B-A3B Instruct

30.5B (3.3B active) · 256K ctx

~21GBat Q4_K_M
CodingAgents & toolsLong context

OpenAI

gpt-oss-20b

20.9B (3.6B active) · 128K ctx

~15GBat MXFP4
ReasoningCodingAgents & tools

Frequently asked questions

How much VRAM do you need to run Qwen3 30B-A3B (2507)?

About 21GB at Q4_K_M with 8K tokens of context: 18.6GB of weights, 0.8GB of KV cache, and 1.2GB of runtime overhead. Higher-precision quants and longer contexts need more; see the table above.

What is the cheapest GPU that can run Qwen3 30B-A3B (2507)?

The smallest card in our database that runs it comfortably is the RTX 4000 Ada (20GB), at Q3_K_M and ~84 tokens/sec.

Is Qwen3 30B-A3B (2507) free for commercial use?

Qwen3 30B-A3B (2507) is released under the Apache 2.0 license. That is a permissive license, so commercial use is allowed.

How long a context does Qwen3 30B-A3B (2507) support?

Up to 262,144 tokens (256K). Long contexts need a lot of extra memory for the KV cache, so most local setups run 8K to 32K.

Sources: Qwen3 30B-A3B Instruct 2507 model card (Hugging Face), Qwen3 30B-A3B Thinking 2507 model card (Hugging Face). Capability score basis: Placed just above the Qwen3 32B anchor from the Artificial Analysis Intelligence Index v3 standing of the 2507 Thinking variant (October 2025).