Can RTX 3090 24GB run gpt-oss 20B?
Measured results for gpt-oss 20B (20.91B MoE, 3.6B active) on RTX 3090 24GB, plus what fits and the commands to run it.
Short answer
Yes. gpt-oss 20B Q8_0 fits fully on RTX 3090 24GB at 16k context, using 12.5 of 23.0 GB usable, and generates about 204.2 tokens/s in our own run. For Q8_0, our bandwidth-only estimate was ~247 tok/s against a measured median of ~204 (21% too optimistic).
Measured generation speed by quantization
| Quant | Generation tok/s | Range | Prompt tok/s | Runs | Origin |
|---|---|---|---|---|---|
| Q8 class (median) | 204.2 | 204.2–204.2 | 5952 | 1 | includes our runs |
| Q4 class (median) | 189.5 | 111.5–282.5 | 6092 | 16 | includes our runs |
| Q4_K_M (our run, depth 0) | 214.3 | 3 repeats | 5857 | 3 | own measurement |
| Q8_0 (our run, depth 0) | 204.2 | 3 repeats | 5952 | 3 | own measurement |
Generation = tg128 (128 generated tokens, short context). Public rows are medians of community llama.cpp results (sources); our own runs use llama-bench b11658 on a rented GPU.
Speed vs context depth measured
Our own llama-bench runs on GeForce RTX 3090 (flash attention on, 3 repeats): generation and prompt speed with the given number of tokens already in the context.
| Quant | Context depth | Generation tok/s | vs depth 0 | Prompt tok/s |
|---|---|---|---|---|
| Q4_K_M | 0 | 214.3 | 100% | 5857 |
| Q4_K_M | 4k | 208.4 | 97% | 5648 |
| Q4_K_M | 16k | 193.8 | 90% | 4546 |
| Q8_0 | 0 | 204.2 | 100% | 5952 |
| Q8_0 | 4k | 198.3 | 97% | 5696 |
| Q8_0 | 16k | 185.2 | 91% | 4522 |
Memory and context limits
Q8_0 weights are 12.1 GB; the 16k f16 KV cache adds 0.41 GB. Largest context that fits in 23.0 GB usable:
| KV cache type | Max context |
|---|---|
| f16 | 128k |
| q8_0 | 128k |
| q4_0 | 128k |
Native limit 128k. Full KV cache table for gpt-oss 20B.
Copy-paste commands
The advisor's pick for this hardware: Q8_0, 16k context.
llama-server -hf unsloth/gpt-oss-20b-GGUF:Q8_0 -c 16384 -fa on --jinja -ngl 99
OLLAMA_CONTEXT_LENGTH=16384 ollama run hf.co/unsloth/gpt-oss-20b-GGUF:Q8_0
FAQ
How fast is gpt-oss 20B on RTX 3090 24GB?
About 204.2 tokens/s for generation (Q8_0 on llama.cpp CUDA, 128-token generation), 5952 tokens/s for prompt processing. Speed drops as the context fills; see the depth table.
How much memory does gpt-oss 20B need on RTX 3090 24GB?
The advisor's pick, Q8_0, is 12.1 GB of weights plus 0.41 GB of KV cache at 16k context (12.5 GB), against 23.0 GB usable.
What is the maximum context for gpt-oss 20B on RTX 3090 24GB?
With an f16 KV cache about 128k tokens, 128k with q8_0 and 128k with q4_0, bounded by the model's own limit.
Related
All models on RTX 3090 24GB · gpt-oss 20B on all hardware · VRAM tier · all measured pages · leaderboard
Compare: RTX 3090 24GB vs RTX 4090 24GB · RTX 3090 24GB vs RTX 5090 32GB · RTX 3090 24GB vs AMD RX 7900 XTX 24GB