ByteDance Seed · bytedance/seed-oss-36b-instruct
| Provider | ByteDance Seed |
|---|---|
| Family | seed |
| Type | llm-chat |
| Status | active |
| Released | 2025-08-20 |
| Updated | 2025-08-26 |
| Parameters | 36.2B |
| Open weights | yes |
Ordered by predicted speed. Bandwidth sets decode rate; memory decides whether it runs at all.
| Device | Memory | Bandwidth | Best quality | Weights | tok/s | Fastest |
|---|---|---|---|---|---|---|
| Cerebras WSE-3 | 44.0 GB | 21000000.0 GB/s | q6 | 27.11 GB | ~542168.8 | ~813253.2 q4 |
| NVIDIA GB200 Grace Blackwell Superchip | 372.0 GB | 16000.0 GB/s | bf16 | 72.3 GB | ~154.9 | ~619.6 q4 |
| AMD Instinct MI355X | 288.0 GB | 8000.0 GB/s | bf16 | 72.3 GB | ~77.5 | ~309.8 q4 |
| NVIDIA B200 | 180.0 GB | 7700.0 GB/s | bf16 | 72.3 GB | ~74.5 | ~298.2 q4 |
| Google TPU7x (Ironwood) | 192.0 GB | 7380.0 GB/s | bf16 | 72.3 GB | ~71.5 | ~285.8 q4 |
| AMD Instinct MI325X | 256.0 GB | 6000.0 GB/s | bf16 | 72.3 GB | ~58.1 | ~232.4 q4 |
| AMD Instinct MI300X | 192.0 GB | 5300.0 GB/s | bf16 | 72.3 GB | ~51.3 | ~205.2 q4 |
| NVIDIA H200 SXM | 141.0 GB | 4800.0 GB/s | bf16 | 72.3 GB | ~46.5 | ~185.9 q4 |
| NVIDIA H100 NVL | 94.0 GB | 3900.0 GB/s | fp8 | 36.15 GB | ~75.5 | ~151.0 q4 |
| NVIDIA H100 SXM | 80.0 GB | 3350.0 GB/s | fp8 | 36.15 GB | ~64.9 | ~129.7 q4 |
| AMD Instinct MI250X | 128.0 GB | 3200.0 GB/s | bf16 | 72.3 GB | ~31.0 | ~123.9 q4 |
| Google TPU v5p | 95.0 GB | 2765.0 GB/s | fp8 | 36.15 GB | ~53.5 | ~107.1 q4 |
| NVIDIA A100 80GB SXM | 80.0 GB | 2039.0 GB/s | fp8 | 36.15 GB | ~39.5 | ~79.0 q4 |
| NVIDIA H100 PCIe | 80.0 GB | 2000.0 GB/s | fp8 | 36.15 GB | ~38.7 | ~77.5 q4 |
| NVIDIA GeForce RTX 5090 | 32.0 GB | 1792.0 GB/s | q5 | 22.59 GB | ~55.5 | ~69.4 q4 |
| Google TPU v6e (Trillium) | 32.0 GB | 1638.0 GB/s | q5 | 22.59 GB | ~50.7 | ~63.4 q4 |
| AMD Instinct MI210 | 64.0 GB | 1600.0 GB/s | fp8 | 36.15 GB | ~31.0 | ~62.0 q4 |
| NVIDIA A100 40GB SXM | 40.0 GB | 1555.0 GB/s | q6 | 27.11 GB | ~40.1 | ~60.2 q4 |
| Google TPU v4 | 32.0 GB | 1200.0 GB/s | q5 | 22.59 GB | ~37.2 | ~46.5 q4 |
| NVIDIA L40S | 48.0 GB | 864.0 GB/s | q6 | 27.11 GB | ~22.3 | ~33.5 q4 |
| Apple M3 Ultra | 512.0 GB | 819.0 GB/s | bf16 | 72.3 GB | ~7.9 | ~31.7 q4 |
| Apple M2 Ultra | 192.0 GB | 800.0 GB/s | bf16 | 72.3 GB | ~7.7 | ~31.0 q4 |
| Apple M4 Max | 128.0 GB | 546.0 GB/s | bf16 | 72.3 GB | ~5.3 | ~21.1 q4 |
| Apple M1 Max | 64.0 GB | 400.0 GB/s | fp8 | 36.15 GB | ~7.7 | ~15.5 q4 |
| Apple M2 Max | 96.0 GB | 400.0 GB/s | fp8 | 36.15 GB | ~7.7 | ~15.5 q4 |
| Apple M3 Max | 128.0 GB | 400.0 GB/s | bf16 | 72.3 GB | ~3.9 | ~15.5 q4 |
| Apple M4 Pro | 64.0 GB | 273.0 GB/s | fp8 | 36.15 GB | ~5.3 | ~10.6 q4 |
| Apple M4 | 32.0 GB | 120.0 GB/s | q5 | 22.59 GB | ~3.7 | ~4.6 q4 |
Taken from the card's availability section.
| Platform | Id on that platform |
|---|---|
| Hugging Face | ByteDance-Seed/Seed-OSS-36B-Instruct |
| Benchmark | Evaluated as | Score | Evidence date | Source kind | |
|---|---|---|---|---|---|
| aa_lcr | Seed-OSS-36B-Instruct | 61.33% | 2026-09-10 evaluated | independent evaluator | source |
| critpt | Seed-OSS-36B-Instruct | 0.0% | 2026-09-10 evaluated | independent evaluator | source |
| gpqa_diamond | Seed-OSS-36B-Instruct | 72.63% | 2026-09-10 evaluated | independent evaluator | source |
This card reports no benchmark scores yet.
| Benchmark | Catalogue standing | Score |
|---|