Nemotron 3 Nano 30B A3B
Trinity Mini
Seed 2.1 Turbo
- Context
- 524Ktokens
- Released
- 1y ago
- Size
- 36Bparameters
Seed-OSS-36B-Instruct is an open-source model released by ByteDance's Seed Team, post-trained from Seed-OSS-36B-Base as part of the Seed-OSS series. Announcing the release, ByteDance said the series is designed for long-context, reasoning, agent and general capabilities, and that training on 12T tokens delivers strong benchmark results at low practical cost. The instruct model is built for work including mathematics, coding, long-text processing, and agentic tasks such as tool-using and issue resolving, with a thinking budget users can set to control reasoning length.
- Language model
- Open weights · 36B
- Reasoning
- Tools
How it compares
Size class
Medium36BAn open-weight model of 25 to 100 billion parameters, for a workstation or a single server GPU.
Open models of a similar size from other labs
Vs. your stack
Add the models you use to see how this one compares.
Sign inAdditional source evidence
- Source model
- bytedance-seed/seed-oss-36b-instruct
- Source family
- seed
- Last updated
- 2025-08-20
- Knowledge cutoff
- 2024-07
- Model context
- 524K
- Model output limit
- 524K
- Open weights
- Yes
- File attachments
- No
- Reasoning
- Yes
- Tool calling
- Yes
- Temperature control
- Yes
Downloadable weights
Complete source record (JSON)
{
"id": "bytedance-seed/seed-oss-36b-instruct",
"name": "Seed-OSS-36B-Instruct",
"description": "Open ByteDance Seed model for long-context reasoning, tool use, and agents",
"family": "seed",
"last_updated": "2025-08-20",
"knowledge": "2024-07",
"attachment": false,
"reasoning": true,
"tool_call": true,
"temperature": true,
"modalities": {
"input": [
"text"
],
"output": [
"text"
]
},
"limit": {
"context": 524288,
"output": 524288
},
"release_date": "2025-08-20",
"open_weights": true,
"license": "Apache 2.0",
"weights": [
{
"label": "Hugging Face",
"url": "https://huggingface.co/ByteDance-Seed/Seed-OSS-36B-Instruct"
}
]
}Frequently asked questions
Running locally
Seed-OSS-36B-Instruct has 36B parameters. Estimated VRAM is 25.6 GB at Q4, 44.5 GB at Q8 and 83.3 GB at FP16. All figures are estimates from the parameter count, covering weights plus about 15% and 0.5 GB for a short context.
No. On a 24 GB, 16 GB, 12 GB or 8 GB GPU, Seed-OSS-36B-Instruct does not fit at Q4, Q8 or FP16 (estimate).
Yes, at Q4, which needs an estimated 25.6 GB. Assuming macOS lets the GPU use about 75% of unified memory, the smallest common Mac memory size that fits Q4 is 36 GB (estimate).
Privacy and data
Yes. Seed-OSS-36B-Instruct has open weights, so you can self-host it for full control of your data.
Capabilities and specs
The context window is 524K tokens (524,288) and the maximum output is 524K tokens (524,288).
No. Input modalities are text only, so Seed-OSS-36B-Instruct cannot read images, audio, video or PDFs.
Yes. Tool calling (function calling) is supported.
Yes. A reasoning mode is supported.
The knowledge cutoff of Seed-OSS-36B-Instruct is 2024-07.
Quality and benchmarks
The closest alternatives, matched on model kind and release date with no comparable price, are MolmoAct 7B-D by Ai2, ERNIE 4.5 21B A3B Thinking by Baidu and Command A Reasoning by Cohere.
MolmoAct 7B-DSeed-OSS-36B-Instruct vs MolmoAct 7B-DERNIE 4.5 21B A3B ThinkingSeed-OSS-36B-Instruct vs ERNIE 4.5 21B A3B ThinkingCommand A ReasoningSeed-OSS-36B-Instruct vs Command A Reasoning
Licensing and access
Open weights. Seed-OSS-36B-Instruct is distributed as open weights under the Apache 2.0 license.
Yes. The license is Apache 2.0, a permissive license that allows commercial use.
The weights are available in the Hugging Face repository ByteDance-Seed/Seed-OSS-36B-Instruct.
Basics
Seed-OSS-36B-Instruct was made by ByteDance and released on August 20, 2025.
Yes. Seed-OSS-36B-Instruct has open weights.
Generated from Vertical Tabs data, updated November 2, 2026.
Discussion
No comments yet
Be the first to share your thoughts.