aliteq.

Type to search every guide, tool and GPU page.

How much VRAM Qwen3-Coder 30B-A3B (MoE) needs, and what it costs to run

Some links on this page are referral links: we may earn a commission, at no cost to you.

Trying to decide what to buy? The best GPU for Qwen3-Coder 30B-A3B (MoE) ↓ · Got a card already? See what your PC can run →

30.5B paramsMoE · 128 experts, 8 active256k max contextapache-2.0

Cheapest way to rent it right now

$0.05/hr

NVIDIA RTX A4000 · Q2_K · spot · ≈ $13/mo at 8h/day

Live price captured Thu, 12 Nov 2026 22:20:09 GMT.It never changes which card is cheapest.

Memory needed, by quantisation

At 16k context. Weights + KV cache + ~0.8 GB overhead. Memory is in binary GB (1,024³ bytes), the unit GPU memory is sold in; Hugging Face and Ollama list downloads in decimal GB, about 7% more.

QuantWeightsKV cacheTotalQuality
F1656.9 GB1.5 GB59.2 GBFull precision. Reference quality, twice the size of Q8 for no practical gain in most chat use.
Q8_030.2 GB1.5 GB32.5 GBEffectively lossless. Use when VRAM is not the constraint.
Q6_K23.3 GB1.5 GB25.6 GBVery close to Q8 at meaningfully less memory. A safe high-quality pick.
Q5_K_M20.2 GB1.5 GB22.5 GBSmall, hard-to-notice quality loss. Good balance.
Q5_019.7 GB1.5 GB22.0 GBOlder-style 5-bit. Q5_K_M is usually the better pick at the same size.
Q4_K_M17.2 GB1.5 GB19.5 GBThe community default. Best quality-per-gigabyte for most people.
Q4_016.2 GB1.5 GB18.5 GBOlder-style 4-bit, measurably worse than Q4_K_M at a similar size. Avoid unless required.
IQ4_XS15.1 GB1.5 GB17.4 GBNewer 4-bit, smaller than Q4_K_M with comparable quality. Needs a recent llama.cpp.
Q3_K_M13.9 GB1.5 GB16.2 GBNoticeable degradation. Use to fit a larger model that would otherwise not run.
Q2_K11.9 GB1.5 GB14.2 GBHeavy degradation. Almost always better to run a smaller model at Q4_K_M instead.
IQ3_XS11.7 GB1.5 GB14.0 GBAggressive. Usually better than a smaller model at Q4, but test before trusting it.

Cards that can run it

Best quantisation each card fits at 16k context, with live rental price where we track one.

GPUVRAMBest fitUses~tok/sRent from
Apple M3 Ultra512 GBF1659 GB——
Apple M5 Ultra512 GBF1659 GB——
Apple M2 Ultra192 GBF1659 GB——
Apple M5 Max (18-core CPU / 40-core GPU)128 GBF1659 GB——
NVIDIA DGX Spark (GB10 Grace Blackwell)128 GBF1659 GB—$0.269/hrVast.ai
Apple M4 Max (16-core CPU / 40-core GPU)128 GBF1659 GB——
AMD Ryzen AI Max+ 395 (Strix Halo, 128 GB)128 GBF1659 GB——
NVIDIA RTX PRO 6000 Blackwell Workstation Edition96 GBF1659 GB—$0.533/hrVast.ai
Apple M2 Max96 GBF1659 GB——
NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition96 GBF1659 GB—$0.333/hrVast.ai
Apple M4 Pro64 GBF16tight59 GB——
NVIDIA RTX 6000 Ada Generation48 GBQ8_032 GB—$0.200/hrVast.ai
NVIDIA RTX PRO 5000 Blackwell48 GBQ8_032 GB—$0.160/hrVast.ai
NVIDIA RTX A600048 GBQ8_032 GB—$0.301/hrVast.ai
AMD Radeon PRO W790048 GBQ8_032 GB——
Apple M4 Max (14-core CPU / 32-core GPU)36 GBQ8_0tight32 GB——
Apple M5 Max (18-core CPU / 32-core GPU)36 GBQ8_0tight32 GB——
NVIDIA RTX PRO 4500 Blackwell32 GBQ6_K26 GB—$0.267/hrVast.ai
NVIDIA GeForce RTX 509032 GBQ6_K26 GB—$0.200/hrVast.ai
NVIDIA RTX 5000 Ada Generation32 GBQ6_K26 GB—$0.265/hrVast.ai
NVIDIA GeForce RTX 3090 Ti24 GBQ5_K_Mtight22 GB—$0.133/hrVast.ai
AMD Radeon RX 7900 XTX24 GBQ5_K_Mtight22 GB——
NVIDIA RTX PRO 4000 Blackwell24 GBQ5_K_Mtight22 GB—$0.133/hrVast.ai
NVIDIA RTX A500024 GBQ5_K_Mtight22 GB—$0.093/hrVast.ai
NVIDIA GeForce RTX 309024 GBQ5_K_Mtight22 GB—$0.080/hrVast.ai
NVIDIA GeForce RTX 409024 GBQ5_K_Mtight22 GB—$0.170/hrNovita AI
NVIDIA RTX 4000 Ada Generation20 GBQ4_K_Mtight20 GB—$0.093/hrVast.ai
AMD Radeon RX 7900 XT20 GBQ4_K_Mtight20 GB——
AMD Radeon RX 9070 XT16 GBQ2_K14 GB——
NVIDIA GeForce RTX 4060 Ti 16GB16 GBQ2_K14 GB—$0.068/hrVast.ai
NVIDIA GeForce RTX 5070 Ti16 GBQ2_K14 GB—$0.120/hrVast.ai
NVIDIA GeForce RTX 5060 Ti 16GB16 GBQ2_K14 GB—$0.072/hrVast.ai
NVIDIA GeForce RTX 4070 Ti SUPER16 GBQ2_K14 GB—$0.067/hrVast.ai
NVIDIA GeForce RTX 408016 GBQ2_K14 GB—$0.183/hrVast.ai
NVIDIA GeForce RTX 4080 SUPER16 GBQ2_K14 GB—$0.132/hrVast.ai
NVIDIA RTX A400016 GBQ2_K14 GB—$0.053/hrVast.ai
AMD Radeon RX 907016 GBQ2_K14 GB——
AMD Radeon RX 9060 XT 16GB16 GBQ2_K14 GB——
NVIDIA GeForce RTX 508016 GBQ2_K14 GB—$0.134/hrVast.ai

The best GPU for Qwen3-Coder 30B-A3B (MoE)

Ranked from live rental prices and computed VRAM fit, not opinion. Every card is judged at Q4_K_M with a 8K context, so the comparison is fair; a card that could only run a crushed version isn't listed.

Best value

NVIDIA GeForce RTX 3090

Best throughput per dollar among cards that fit.

Cheapest that runs it

NVIDIA GeForce RTX 3090

Lowest hourly rental that fits it — $0.080/hr, Q4_K_M.

No compromise

—

Most headroom and highest throughput.

Every card that runs it, ranked

Sorted by tokens/sec per rental dollar. Throughput is a bandwidth-derived estimate.

GPUVRAMVRAM used~tok/sCheapest rentaltok/s per $
NVIDIA RTX PRO 6000 Blackwell Workstation Edition96 GB19 GB20%—$0.533—
NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition96 GB19 GB20%—$0.333—
NVIDIA RTX 6000 Ada Generation48 GB19 GB39%—$0.200—
NVIDIA RTX PRO 5000 Blackwell48 GB19 GB39%—$0.160—
NVIDIA RTX A600048 GB19 GB39%—$0.301—
AMD Radeon PRO W790048 GB19 GB39%———
NVIDIA RTX PRO 4500 Blackwell32 GB19 GB59%—$0.267—
NVIDIA GeForce RTX 509032 GB19 GB59%—$0.200—
NVIDIA RTX 5000 Ada Generation32 GB19 GB59%—$0.265—
NVIDIA GeForce RTX 3090 Ti24 GB19 GB78%—$0.133—
AMD Radeon RX 7900 XTX24 GB19 GB78%———
NVIDIA RTX PRO 4000 Blackwell24 GB19 GB78%—$0.133—

The value column ranks renting, because that's what we can price precisely (prices captured Thu, 12 Nov 2026 22:20:09 UTC). Buying a card? The fit and throughput columns are what you need; we don't publish a purchase-price ranking because we don't hold verified street prices. Tokens/sec is an estimate, not a benchmark. This mixture-of-experts model has no known active-parameter count, so throughput is left out rather than guessed. Got a card already? See everything your PC can run.

qwen3-coder-30b-a3b

How these numbers are produced

  • Model shape comes from the model's own config.json — 48 layers, 32 attention heads, 4 KV heads.
  • Weight size is parameters × effective bits-per-weight. Those constants are checked against real published quantised file sizes — median error 0.7% across 40 measurements.
  • KV cache is 2 × layers × KV-heads × head-dim × context × 2 bytes. Using KV-heads rather than attention heads is what makes this correct for grouped-query attention; treating a GQA model as multi-head overstates the cache by up to 8×.
  • Tokens/sec is an estimate, not a benchmark. Generation is memory-bandwidth-bound, so this is bandwidth ÷ weight-bytes derated to 75%. Real throughput depends on your runtime, batch size and kernels. We don't run our own hardware tests — see our editorial policy.
  • This is a mixture-of-experts model. Generation reads only the experts active for each token (8 of 128), so throughput is scored on active parameters, not the full 31B. Memory is the opposite: every expert must still be resident in VRAM, so the figures above use the full weight set.Derived from architecture (config.json); MoE routes ~3.3B active params per token.
  • Some memory-bandwidth figures are not yet independently verified.Where that's the case we show no tokens/sec at all rather than a number we can't stand behind.
  • Rental prices are pulled hourly from provider APIs (last updated Thu, 12 Nov 2026 22:20:09 GMT). Spot/interruptible pricing can change or vanish without notice.

Try it interactively

Drag the context slider and watch the KV cache fill the card

Running Qwen3-Coder 30B-A3B (MoE): common questions

How much VRAM do you need to run Qwen3-Coder 30B-A3B (MoE)?
At Q4_K_M and at 8k context, Qwen3-Coder 30B-A3B (MoE) needs about 18.8 GB — 17.2 GB of weights plus 0.8 GB of KV cache and ~0.8 GB overhead. So you want a card with at least that much free memory; the KV cache grows if you use longer context.
Can an RTX 4090 run Qwen3-Coder 30B-A3B (MoE)?
Yes. At Q4_K_M it uses about 18.8 GB of the card's 24 GB (78%).
Can an RTX 3090 run Qwen3-Coder 30B-A3B (MoE)?
Yes. At Q4_K_M it uses about 18.8 GB of the card's 24 GB (78%).
Can an RTX 4060 Ti 16GB run Qwen3-Coder 30B-A3B (MoE)?
Not at Q4_K_M — Qwen3-Coder 30B-A3B (MoE) needs more than the card's 16 GB at that quality and 8k context. You'd have to drop to a lower quantisation (worse quality), shorten the context, or use a bigger card or two cards.
What's the cheapest way to run Qwen3-Coder 30B-A3B (MoE)?
Among cards that fit it at Q4_K_M, the cheapest to rent right now is the NVIDIA GeForce RTX 3090 at about $0.080/hour. Whether renting or buying is cheaper overall depends on how many hours a day you'll actually use it.

Worth reading before you buy

Other models