Can the GeForce RTX 5070 run gpt-oss-20b?
Only with CPU offload at Q4_K_M.
gpt-oss-20b is 21B parameters (MoE; all experts resident). At Q4_K_M the weights are about 11.7 GiB; with the 1.5 GiB runtime reserve that is 13.2 GiB against the card's 12GB, about 1.2 GiB more than the card has.
All three quants
12GB card| Quant | Weights | With reserve | Verdict | Headroom |
|---|---|---|---|---|
| Q4_K_Mthe everyday quant | 11.7 GiB | 13.2 GiB | Only with CPU offload | short 1.2 GiB |
| Q8_0near-lossless | 20.7 GiB | 22.2 GiB | Only with CPU offload | short 10.2 GiB |
| FP16full weights | 39.1 GiB | 40.6 GiB | Does not fit | short 28.6 GiB |
Verdict rules: fits comfortably when weights plus reserve sit within 85% of VRAM; tight when they fit with little left for context; CPU offload when they exceed VRAM by up to 2x (runs, slowly, with layers in system RAM); no beyond that. The KV cache per token depends on the model's architecture and is not modelled here.
The numbers for this pairing
At Q4_K_M the weights take about 11.7 GiB, so with the 1.5 GiB runtime reserve it needs 13.2 GB: 1.2 GB more than the card's 12 GB.
No version of it fits entirely in this card's memory, so it would lean on system RAM and run far slower.
It is a mixture-of-experts model: about 3.6B of its 21B parameters work on each token, which helps speed, but all of them still have to fit in memory.
The cheapest card on our list that runs it comfortably at Q4 is the Radeon RX 9060 XT 16GB at about $530, against about $801 for this card.
At about $801, this card costs roughly $67 per GB of VRAM.
Running it at its full 250 W board power for 4 hours a day would use about 30 kWh a month, about $5.52 at the US average residential rate of 18.4 cents per kWh (2026-05); real use sits below that ceiling.
Prices are the lowest current listings our tracker saw on Oct 7, 2026, since averages include premium variants; launch prices are marked where we have no listing.
Not this card. These fit it comfortably at Q4.
- GeForce RTX 5090 32GB · headroom 18.8 GiB
- GeForce RTX 3090 (used) 24GB · headroom 10.8 GiB
- GeForce RTX 4090 (used) 24GB · headroom 10.8 GiB
- Radeon RX 7900 XTX 24GB · headroom 10.8 GiB
What the GeForce RTX 5070 does run well → · gpt-oss-20b on every card →
About the model
gpt-oss-20b by OpenAI: 131,072-token native context, text, released 2025 under the Apache-2.0. Model card → · Hosted pricing on PicksByModel →