Leaderboard / ByteDance Seed

Seed-OSS 36B Instruct

Name: Seed-OSS 36B Instruct
Author: ByteDance Seed

ByteDance Seed · Seed-OSS · released 2025-08-20 · data checked 2026-11-01

Seed-OSS 36B Instruct is a 36.2B-parameter dense model from ByteDance Seed (China), released 2025-08-20 under the Apache-2.0 license. It scores 60.3 on the Small LM Index, #20 of 34 medium models. It needs about 25.3 GB of memory at 4-bit with an 8K-token context.

Thinking budget can be set per request.

Small LM Index

60.3

#20 of 34 scored medium models

Likely 56.7–63.8 · from 6 of 6 core benchmarks

Parameters

36.2B

Dense

Price per 1M tokens

—

No hosted API on OpenRouter; self-host

Context window

512K

tokens

Benchmarks

Each bar shows this model's score. The tick marks the best medium model on that benchmark. Sources: 9 reported by ByteDance Seed. How we label sources.

MMLU-Pro Lab-reported Knowledge82.7

Class best: Qwen3.6 27B 86.2

GPQA Diamond Lab-reported Knowledge71.4

Class best: Nanbeige4.3 33B 89.5

Humanity's Last Exam Lab-reported Reasoning10.1

Class best: Qwen3.8 27B 30.8

AIME Lab-reported Math · AIME 202584.7

Class best: Ternary Bonsai 2 27B 95.8

LiveCodeBench Lab-reported Coding · LiveCodeBench v667.4

Class best: Qwen3.8 27B 90.3

SWE-bench Verified Lab-reported Coding · OpenHands56.0

Class best: Qwen3.6 27B 77.2

MMLU Lab-reported Knowledge87.4

Class best: EXAONE 4.0 32B 92.3

Multilingual MMLU Lab-reported Multilingual · MMMLU78.4

Class best: EXAONE 4.0 32B 85.6

IFEval Lab-reported Instruction following85.8

Class best: LFM2.5 24B-A2B 96.8

Specifications

Lab: ByteDance Seed (CN · China)
Architecture: Dense transformer
Total parameters: 36.2B
Active parameters: All (dense)
Size class: Medium
Context window: 512K tokens
Input modalities: text
Reasoning: Reasoning (thinking)
License: Apache-2.0
Openness: Open weights
Memory to run: ~25.3 GB at 4-bit, 8K context (more sizes)
Hugging Face: 40.7K downloads (30 days) · 506 likes

Weights on Hugging Face ↗

Compare with

Seed-OSS 36B Instruct vs Qwen3.6 35B-A3B (35B, index 72.2)
Seed-OSS 36B Instruct vs Qwen3.5 35B-A3B (35B, index 69.9)
Seed-OSS 36B Instruct vs Edge0 35B-A3B (preview) (34.7B, index 62.8)
Seed-OSS 36B Instruct vs Hermes 4.3 36B (36.2B, index 51.4)
Seed-OSS 36B Instruct vs MiniCPM-V 4.7 35B-A3B (35.2B, index 54.5)

Sources

https://huggingface.co/ByteDance-Seed/Seed-OSS-36B-Instruct

Something wrong here? Report it.

Run it locally

Uses unsloth/Seed-OSS-36B-Instruct-GGUF (Q4_K_M). Swap the tag for another size: BF16, IQ4_NL, IQ4_XS, Q2_K, Q2_K_L, Q3_K_M, Q3_K_S, Q4_0….

Ollama

ollama run hf.co/unsloth/Seed-OSS-36B-Instruct-GGUF:Q4_K_M

llama.cpp

llama-server -hf unsloth/Seed-OSS-36B-Instruct-GGUF:Q4_K_M

Then open http://localhost:8080 for a chat window.

MLX (Mac)

pip install -U mlx-lm
mlx_lm.chat --model mlx-community/Seed-OSS-36B-Instruct-4bit

vLLM (GPU server)

vllm serve ByteDance-Seed/Seed-OSS-36B-Instruct

Full-precision weights, for a server GPU.

Open in LM Studio

Memory needed: weights, context cache and runtime buffers.

Quantization	4K context	32K context	256K context
16-bit F16	75 GB	82 GB	142 GB
8-bit Q8_0	41 GB	48 GB	108 GB
4-bit Q4_K_M	24 GB	32 GB	92 GB

Check it against your GPU, Mac or phone →