Best GPU for gpt-oss-20b Locally
Real-time prices and hardware recommendations updated for August 2026.
gpt-oss-20b is a dense model: all 20B parameters activate on every token, so generation speed is bound by how fast your card can stream the full weights.
To run gpt-oss-20b locally you need roughly 14.1 GB of VRAM at Q4_K_M quantization with a 32k token context. The best-value card that fits is the GeForce RTX 5060 Ti 16GB (16 GB), which should generate around 12 tokens per second.
Adjust Context Length
Slide to popular sizes or type any value manually.
Q3 weights (~8.6GB) plus 32k KV cache reach ~11GB. 12GB cards cover the full context window with headroom.
Q4 weights (~11.5GB) plus 32k KV cache reach ~14GB. 16GB cards are the ideal mainstream setup.
Q8 weights (~23GB) plus 32k KV cache exceed 24GB. 32GB+ cards deliver near-lossless precision without VRAM constraint.
Other models with the same VRAM requirement
Because gpt-oss-20b's weights fit a 16 GB card, other models of a similar size run on the same GPU. Generation speed varies — mixture-of-experts models are faster, dense models slower — but any of these load in the same VRAM:
Optimizing Setup for gpt-oss-20b
Quantization Recommendations
For daily coding and reasoning tasks, Q4_K_M (4-bit quantization) offers the best balance of quality and memory efficiency — it reduces memory requirements by over 70% with minimal quality loss compared to FP16. Q8 and higher presets preserve more fidelity at the cost of significantly higher VRAM usage, which may force layer offloading and hurt throughput.
Recommended Local Software
We recommend using Ollama as the primary runner for local inference due to its automated GPU model splitting and context cache optimizations. For advanced fine-tuning or quantization splits, llama.cpp with Flash Attention compiled natively provides the best granular control.
Running gpt-oss-20b locally — FAQ
How much VRAM do I need to run gpt-oss-20b?
At a 32k context with KV cache quantization on, gpt-oss-20b needs about 14.1 GB of VRAM at Q4_K_M — the quantization most people should use. Dropping to Q3_K_M brings that down to roughly 11.2 GB at some quality cost, while Q8_0 needs about 25.6 GB for the best quality this model can give.
What size graphics card does gpt-oss-20b fit on?
gpt-oss-20b needs about 14.1 GB at Q4_K_M, so a 16 GB card is the smallest common size that holds it entirely in VRAM. Anything smaller has to offload layers to system RAM, which typically costs you most of your generation speed.
Which quantization should I use for gpt-oss-20b?
Use Q4_K_M unless you have VRAM to spare. It needs about 14.1 GB and loses very little quality against full precision. Q8_0 needs about 25.6 GB for a quality gain most people cannot detect in everyday coding and chat. Spend spare VRAM on a longer context instead.
How does context length affect the VRAM gpt-oss-20b needs?
Model weights are fixed, but the KV cache grows linearly with context. For gpt-oss-20b at a 32k context the cache is about 2.6 GB; doubling to 64k takes it to roughly 5.1 GB. Turning KV cache quantization off doubles those figures again.
Can the same GPU run other models similar to gpt-oss-20b?
Yes. gpt-oss-20b needs about 14.1 GB at Q4_K_M, and any model whose weights fit the same card runs on it — including InternLM2.5-20B (20B), Codestral 22B (22.2B), DiffusionGemma 26B-A4B (25.2B). The weights fit the same GPU; generation speed varies (mixture-of-experts models are faster, dense models slower).
▸How token speeds are estimated
Two metrics are shown per GPU: Read tok/s (how fast the model ingests your prompt) and Decode tok/s (how fast it streams tokens back). They model fundamentally different bottlenecks.
Read (Prefill)
The prompt is processed in one parallel pass. This is compute-bound: it saturates the GPU's tensor cores.
Decode (Generation)
Each new token requires loading the entire model's active weights from VRAM. This is memory-bandwidth-bound: the GPU stalls waiting for data, not computing.
Weights = (activeParams × bits ÷ 8) × 1.15 overhead. KV cache per step = activeParams × multiplier × contextK.
Architecture utilization factors
Left: decode factor — Right: read factor
Data sources
TFLOPS and memory bandwidth are read from the GPU database. When missing, bandwidth falls back to a hardcoded dictionary.
Limitations
These are analytical estimates, not benchmark results. Use them as a relative comparison, not an absolute performance guarantee.