Baltor Get started

Models

Seed-OSS-36B-Instruct

By ByteDance-Seed. 36.2 billion parameters, a context window of 524,288 tokens and the licence apache-2.0.

Facts

Released
2025-08-20 the day the repository was first published on Hugging Face Hugging Face, read
Licence
apache-2.0 the licence the model card declares Hugging Face, read
Open weights
Yes the weights are published in this Hugging Face repository Hugging Face, read
Parameters
36.2 billion counted from the safetensors weight files Hugging Face, read
Active parameters
Unknown
Knowledge cutoff
Unknown
Context window
  • 524,288 tokens the maximum position embeddings in the model configuration model configuration, read
Longest output
Unknown
Tool calling
Unknown
Structured output
Unknown
Reasoning controls
Unknown
Inputs and outputs
text in, text out Hugging Face, read
Good for
No source names a use.

Prices

Prices in US dollars per million tokens, as each source lists them
ProviderInputOutputContextAs ofSource
SiliconFlow direct0.210 USD0.570 USD262,000 older than 30 days; check the providermodels.dev, read
SiliconFlow (China) direct0.210 USD0.570 USD262,000 older than 30 days; check the providermodels.dev, read

Run it on your own hardware

Weights are the sizes of the files a source lists, or an estimate from the parameter count where none does. Memory is an estimate: weights plus KV cache plus 512 MiB and 5 percent of the weights for runtime buffers. Check it against your hardware.

QuantizationWeightsMemory at 8,192 tokensMemory at 32,768 tokens
UD-IQ1_S7.9 GiB10.8 GiB16.8 GiB
UD-IQ1_M8.5 GiB11.4 GiB17.4 GiB
UD-IQ2_XXS9.5 GiB12.4 GiB18.4 GiB
UD-IQ2_M11.9 GiB14.9 GiB20.9 GiB
Q2_K12.7 GiB15.8 GiB21.8 GiB
Q2_K_L12.8 GiB16.0 GiB22.0 GiB
UD-Q2_K_XL13.1 GiB16.2 GiB22.2 GiB
UD-IQ3_XXS13.3 GiB16.4 GiB22.4 GiB
Q3_K_S14.8 GiB18.0 GiB24.0 GiB
Q3_K_M16.4 GiB19.7 GiB25.7 GiB
UD-Q3_K_XL16.8 GiB20.1 GiB26.1 GiB
IQ4_XS18.2 GiB21.6 GiB27.6 GiB
IQ4_NL19.2 GiB22.6 GiB28.6 GiB
Q4_019.2 GiB22.7 GiB28.7 GiB
Q4_K_S19.3 GiB22.7 GiB28.7 GiB
Q4_K_M20.3 GiB23.8 GiB29.8 GiB
UD-Q4_K_XL20.5 GiB24.0 GiB30.0 GiB
Q4_121.2 GiB24.8 GiB30.8 GiB
Q5_K_S23.3 GiB26.9 GiB32.9 GiB
Q5_K_M23.8 GiB27.5 GiB33.5 GiB
UD-Q5_K_XL23.9 GiB27.6 GiB33.6 GiB
Q6_K27.6 GiB31.5 GiB37.5 GiB
UD-Q6_K_XL29.7 GiB33.6 GiB39.6 GiB
Q8_035.8 GiB40.1 GiB46.1 GiB
UD-Q8_K_XL40.0 GiB44.5 GiB50.5 GiB
BF1667.3 GiB73.2 GiB79.2 GiB

Start it with a local runtime

  • Ollama
    ollama run hf.co/unsloth/Seed-OSS-36B-Instruct-GGUF:Q4_0
  • LM Studio
    lms get unsloth/Seed-OSS-36B-Instruct-GGUF
    lms server start
  • llama.cpp server
    llama-server -hf unsloth/Seed-OSS-36B-Instruct-GGUF:Q4_0 --jinja
  • vLLM
    vllm serve ByteDance-Seed/Seed-OSS-36B-Instruct
  • SGLang
    sglang serve --model-path ByteDance-Seed/Seed-OSS-36B-Instruct --port 30000

Use it from your harness

Set up for SiliconFlow with the model ByteDance-Seed/Seed-OSS-36B-Instruct. Each endpoint page has the same setup for its own address.

OpenCode

Put this in opencode.json in your project folder:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "siliconflow": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "SiliconFlow",
      "options": {
        "baseURL": "https://api.siliconflow.com/v1",
        "apiKey": "{env:SILICONFLOW_API_KEY}"
      },
      "models": {
        "ByteDance-Seed/Seed-OSS-36B-Instruct": {
          "name": "Seed-OSS-36B-Instruct"
        }
      }
    }
  }
}
  • OpenCode reads any OpenAI-compatible address through the @ai-sdk/openai-compatible package, and an address that speaks the Responses API through @ai-sdk/openai.

From OpenCode documentation, read .

Pi

Put this in ~/.pi/agent/models.json:

{
  "providers": {
    "siliconflow": {
      "baseUrl": "https://api.siliconflow.com/v1",
      "api": "openai-completions",
      "apiKey": "$SILICONFLOW_API_KEY",
      "models": [
        {
          "id": "ByteDance-Seed/Seed-OSS-36B-Instruct"
        }
      ]
    }
  }
}
  • The apiKey field can name an environment variable as $NAME.

From Pi documentation, read .

Codex

Codex speaks only the Responses API, and SiliconFlow documents no Responses address. A gateway that offers one can sit in between.

  • Codex speaks the Responses API only: responses is the one supported wire API of a custom provider. Ollama and LM Studio are built in and start with --oss.

From Codex documentation, read .

Claude Code

Claude Code sends Anthropic Messages requests, and SiliconFlow documents no such address. A gateway that translates to that API can sit in between.

  • Claude Code sends Anthropic Messages requests to ANTHROPIC_BASE_URL. Anthropic says it does not support routing Claude Code to models other than Claude through any gateway, so some features may not work with another model.

From Claude Code documentation, read .

Other ways to reach it

Published results

These are results other people published. Baltor did not run them and does not rank models by them.