RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can Seed OSS 36B Instruct run on the 2x RTX 3090?

Yes. At 8K context the best fit is Q8_0, using about 38.5 GB of the 45.6 GB usable on the 2x RTX 3090. Estimated speed ceiling: ~24 tokens/sec.

Memory by quant and context

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K) •14.4 GB Fits15.4 GB Fits21.4 GB Fits45.4 GB Tight69
3-bit (Q3_K_M) •18.1 GB Fits19.1 GB Fits25.1 GB Fits49.1 GB No53
4-bit (Q4_K_M) •22.0 GB Fits23.0 GB Fits29.0 GB Fits53.0 GB No43
5-bit (Q5_K_M) •25.5 GB Fits26.5 GB Fits32.5 GB Fits56.5 GB No37
6-bit (Q6_K) •29.3 GB Fits30.3 GB Fits36.3 GB Fits60.3 GB No32
8-bit (Q8_0) •37.5 GB Fits38.5 GB Fits44.5 GB Tight68.5 GB No24
16-bit (F16/BF16) •69.0 GB No70.0 GB No76.0 GB No100 GB No—

• real GGUF file size from unsloth/Seed-OSS-36B-Instruct-GGUF. Speed is a bandwidth ceiling (936 GB/s), not a benchmark. Methodology.

More

Seed OSS 36B Instruct on every GPU and Mac · Every model that runs on the 2x RTX 3090 · Try other settings in the calculator

Data: Hugging Face Hub (ByteDance-Seed/Seed-OSS-36B-Instruct), NVIDIA specs. Data as of 2026-11-17.