aliteq.

Type to search every guide, tool and GPU page.

How much VRAM gpt-oss 20B (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 gpt-oss 20B (MoE) ↓ · Got a card already? See what your PC can run →

20.9B paramsMoE · 32 experts, 4 active128k max contextapache-2.0

Cheapest way to rent it right now

$0.03/hr

NVIDIA L4 · MXFP4 · spot · ≈ $8/mo at 8h/day

Live price captured Thu, 19 Nov 2026 18:20:07 GMT.It never changes which card is cheapest.

Memory needed, by quantisation

At 16k context. Weights + KV cache + ~0.8 GB overhead. gpt-oss 20B (MoE) ships in MXFP4, so its own release leads and no bigger copy is listed: it would only store the same numbers in more bytes. 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
MXFP4official12.8 GB0.8 GB14.3 GBThe model's own release (MXFP4, a 13.8 GB download): the best quality there is. A bigger copy only stores the same numbers in more bytes. Its 4-bit GGUF builds keep the 4-bit experts as they are, so they come out about this size too.

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 GBMXFP414 GB——
Apple M5 Ultra512 GBMXFP414 GB——
Apple M2 Ultra192 GBMXFP414 GB——
Apple M5 Max (18-core CPU / 40-core GPU)128 GBMXFP414 GB——
NVIDIA DGX Spark (GB10 Grace Blackwell)128 GBMXFP414 GB—$0.269/hrVast.ai
Apple M4 Max (16-core CPU / 40-core GPU)128 GBMXFP414 GB——
AMD Ryzen AI Max+ 395 (Strix Halo, 128 GB)128 GBMXFP414 GB——
NVIDIA RTX PRO 6000 Blackwell Workstation Edition96 GBMXFP414 GB—$0.400/hrVast.ai
Apple M2 Max96 GBMXFP414 GB——
NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition96 GBMXFP414 GB—$0.333/hrVast.ai
Apple M4 Pro64 GBMXFP414 GB——
NVIDIA RTX 6000 Ada Generation48 GBMXFP414 GB—$0.269/hrVast.ai
NVIDIA RTX PRO 5000 Blackwell48 GBMXFP414 GB—$0.163/hrVast.ai
NVIDIA RTX A600048 GBMXFP414 GB—$0.200/hrVast.ai
AMD Radeon PRO W790048 GBMXFP414 GB——
Apple M4 Max (14-core CPU / 32-core GPU)36 GBMXFP414 GB——
Apple M5 Max (18-core CPU / 32-core GPU)36 GBMXFP414 GB——
NVIDIA RTX PRO 4500 Blackwell32 GBMXFP414 GB—$0.200/hrVast.ai
NVIDIA GeForce RTX 509032 GBMXFP414 GB—$0.200/hrVast.ai
NVIDIA RTX 5000 Ada Generation32 GBMXFP414 GB—$0.265/hrVast.ai
NVIDIA GeForce RTX 3090 Ti24 GBMXFP414 GB—$0.173/hrVast.ai
AMD Radeon RX 7900 XTX24 GBMXFP414 GB——
NVIDIA RTX PRO 4000 Blackwell24 GBMXFP414 GB—$0.100/hrVast.ai
NVIDIA RTX A500024 GBMXFP414 GB—$0.200/hrVast.ai
NVIDIA GeForce RTX 309024 GBMXFP414 GB—$0.082/hrVast.ai
NVIDIA GeForce RTX 409024 GBMXFP414 GB—$0.170/hrNovita AI
NVIDIA RTX 4000 Ada Generation20 GBMXFP414 GB—$0.107/hrVast.ai
AMD Radeon RX 7900 XT20 GBMXFP414 GB——
AMD Radeon RX 9070 XT16 GBMXFP414 GB——
NVIDIA GeForce RTX 4060 Ti 16GB16 GBMXFP414 GB—$0.056/hrVast.ai
NVIDIA GeForce RTX 5070 Ti16 GBMXFP414 GB—$0.120/hrVast.ai
NVIDIA GeForce RTX 5060 Ti 16GB16 GBMXFP414 GB—$0.042/hrVast.ai
NVIDIA GeForce RTX 4070 Ti SUPER16 GBMXFP414 GB—$0.093/hrVast.ai
NVIDIA GeForce RTX 408016 GBMXFP414 GB—$0.120/hrVast.ai
NVIDIA GeForce RTX 4080 SUPER16 GBMXFP414 GB—$0.148/hrVast.ai
NVIDIA RTX A400016 GBMXFP414 GB—$0.067/hrVast.ai
AMD Radeon RX 907016 GBMXFP414 GB——
AMD Radeon RX 9060 XT 16GB16 GBMXFP414 GB——
NVIDIA GeForce RTX 508016 GBMXFP414 GB—$0.173/hrVast.ai

The best GPU for gpt-oss 20B (MoE)

Ranked from live rental prices and computed VRAM fit, not opinion. Every card is judged at its own MXFP4 release 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 5060 Ti 16GB

Best throughput per dollar among cards that fit.

Cheapest that runs it

NVIDIA GeForce RTX 5060 Ti 16GB

Lowest hourly rental that fits it — $0.042/hr, MXFP4.

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 GB14 GB15%—$0.400—
NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition96 GB14 GB15%—$0.333—
NVIDIA RTX 6000 Ada Generation48 GB14 GB29%—$0.269—
NVIDIA RTX PRO 5000 Blackwell48 GB14 GB29%—$0.163—
NVIDIA RTX A600048 GB14 GB29%—$0.200—
AMD Radeon PRO W790048 GB14 GB29%———
NVIDIA RTX PRO 4500 Blackwell32 GB14 GB44%—$0.200—
NVIDIA GeForce RTX 509032 GB14 GB44%—$0.200—
NVIDIA RTX 5000 Ada Generation32 GB14 GB44%—$0.265—
NVIDIA GeForce RTX 3090 Ti24 GB14 GB58%—$0.173—
AMD Radeon RX 7900 XTX24 GB14 GB58%———
NVIDIA RTX PRO 4000 Blackwell24 GB14 GB58%—$0.100—

The value column ranks renting, because that's what we can price precisely (prices captured Thu, 19 Nov 2026 18:20:07 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. Got a card already? See everything your PC can run.

gpt-oss-20b

How these numbers are produced

  • Model shape comes from the model's own config.json — 24 layers, 64 attention heads, 8 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 (4 of 32), so throughput is scored on ~3.6B active parameters, not the full 21B. Memory is the opposite: every expert must still be resident in VRAM, so the figures above use the full weight set.Active = 3.6B, vendor figure: OpenAI model card (huggingface.co/openai/gpt-oss-20b, read 29 Oct 2026): "21B parameters with 3.6B active parameters". params = HF safetensors total. Also set in workers/prices/src/models.ts seeds so the daily ingest keeps it.
  • 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, 19 Nov 2026 18:20:07 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 gpt-oss 20B (MoE): common questions

How much VRAM do you need to run gpt-oss 20B (MoE)?
In its own MXFP4 release and at 8k context, gpt-oss 20B (MoE) needs about 14.0 GB — 12.8 GB of weights plus 0.4 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 gpt-oss 20B (MoE)?
Yes. In its own MXFP4 release it uses about 14.0 GB of the card's 24 GB (58%).
Can an RTX 3090 run gpt-oss 20B (MoE)?
Yes. In its own MXFP4 release it uses about 14.0 GB of the card's 24 GB (58%).
Can an RTX 4060 Ti 16GB run gpt-oss 20B (MoE)?
Yes. In its own MXFP4 release it uses about 14.0 GB of the card's 16 GB (87%).
What's the cheapest way to run gpt-oss 20B (MoE)?
Among cards that fit it in its own MXFP4 release, the cheapest to rent right now is the NVIDIA GeForce RTX 5060 Ti 16GB at about $0.042/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