Seed OSS 36B Base — Hardware Requirements & GPU Compatibility
Chat
Seed-OSS-36B-Base is ByteDance's 36-billion-parameter open-source base language model, a dense transformer using GQA attention, RMSNorm, and SwiGLU, designed as a foundation for long-context, reasoning, and agentic downstream use. It is a pretrained model, not instruction-tuned, though this particular checkpoint's pretraining mix includes synthetic instruction data (ByteDance separately released a Base-woSyn variant trained without it, for researchers who want a foundation model unaffected by synthetic instruction data). Despite training on only 12 trillion tokens, it scores competitively against larger open pretrained models on knowledge, reasoning, and math benchmarks. At 36B parameters it needs a multi-GPU workstation or heavy quantization to run locally. Context length is 524,288 tokens, trained natively rather than extended after the fact. It is released under the Apache 2.0 license, permitting unrestricted commercial and research use, and was published in August 2025 alongside the instruction-tuned Seed-OSS-36B-Instruct.
Decent on a 24 GB GPU
Runs from 16 GB (IQ2_M · 16K)Best on 24 GB (Q4_K_M · ~29 tok/s · 8K)Macs from 16 GB (IQ2_XXS · ~13 tok/s)
Apache 2.0
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Seed OSS 36B Base
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
ollama run hf.co/mradermacher/Seed-OSS-36B-Base-i1-GGUF:Q4_K_MQ4_K_M · needs ~24.1 GB · 16K context
Download: Seed-OSS-36B-Base.i1-Q4_K_M.gguf21.76 GBmradermacher
Ways to runOther ways
each opens step-by-stepRun in cloud
Your own private copy · Q4_K_M · 32 GB GPU
- The easiest way to try this model
- Running in a few minutes, easy setup
- Chat right here or connect any app
$5+ bonus for new users
Runs on RTX 5090 (32 GB) (7 GB headroom) · Q4_K_M
- Generation speed
- ~52 tok/s
- generation speed
- Cost per 1M output tokens
- $2.64
- per 1M output tokens
How Much VRAM Does Seed OSS 36B Base Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 15.9 GB | 20 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 32K context · ~41 tok/s | |
| 16.4 GB | 20.4 GB | 3-bit small — 3-bit small quantization | Fits: 32K context · ~40 tok/s | |
| 18.2 GB | 22.2 GB | 3-bit medium — 3-bit medium quantization | Fits: 32K context · ~36 tok/s | |
| 18.6 GB | 22.7 GB | 4-bit · legacy — 4-bit legacy quantization | Fits: 32K context · ~35 tok/s | |
| 22.3 GB | 26.3 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 8K context · ~29 tok/s | |
| 26.3 GB | 30.4 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Does not fit: needs 28.2 GB | |
| 30.4 GB | 34.4 GB | 6-bit, very good — 6-bit quantization, very good quality | Does not fit: needs 32.3 GB | |
| 36.7 GB | 40.8 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Does not fit: needs 38.6 GB |
Selected Q4_K_M
4-bit medium quantization24.1GBat 16K context
- Weights
- 21.7 GB
- KV cache
- 2.2 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · fits up to 8K context · ~29 tok/s
≥50 tok/s needs ≥1.71 TB/s · RTX 5090 ~52
Or .
Which Hardware Can Run Seed OSS 36B Base?
Q4_K_M · 22.3 GBSeed OSS 36B Base (Q4_K_M) requires 22.3 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 29+ GB is recommended. Using the full 524K context window can add up to 68.4 GB, bringing total usage to 90.7 GB. 7 GPUs can run it, including NVIDIA GeForce RTX 5090. 100 devices with unified memory can run Seed OSS 36B Base, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio M5 Max (36 GB).
20 GB
Q4_K_M · best fit IQ3_M (18.7 GB)
1 GPU
RX 7900 XT~29Simulate output on RX 7900 XT32–40 GB
fits · 7.9 GB free · up to 64K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~52Simulate output on RTX 5090M5 Max36 GB~15Simulate output on Mac Studio M5 Max (36 GB)M532 GB~5Simulate output on MacBook Air 13" M5 (32 GB)Orin~6Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~4Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
How good is Seed OSS 36B Base?
llmrun Score blends public benchmark boards into one number; 100 = the average open model.
Not enough data yet for an llmrun Score.
A Score needs results on enough independent boards. How the Composite Score is built
Benchmarks
This is the single board showing results for this model; badges show who ran the test.
- MMLU-Pro
TIGER-LabIndependent
Independent65.1%
#47/146 open
Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Qwen3.6 35B A3BAlibaba | 106.8 | 22 GB | ~151 | same size |
| Sage A1.5InternScience | — | 22 GB | ~151 | same size |
| MiniCPM V 4.7 35B A3BOpenBMB | — | 22.1 GB | ~169 | same size |
| Ornith 1.5 35B A3Bornith-ai | — | 22 GB | ~151 | same size |
| Agents A1InternScience | — | 21.5 GB | ~171 | same size |
| Intern S2 PreviewInternLM | — | 73.2 GB | —needs > 24 GB | same size |
Fine-tunes & variants(1)
Specifications
- Publisher
- ByteDance-Seed
- Family
- Seed-OSS
- Parameters
- 36.2B
- Architecture
- SeedOssForCausalLM
- Context Length
- 524,288 tokens
- Vocabulary Size
- 155,136
- Release Date
- 2025-08-20
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does Seed OSS 36B Base need?
Seed OSS 36B Base requires 22.3 GB of VRAM at Q4_K_M, or 72.9 GB at BF16. Full 524K context adds up to 68.4 GB (90.7 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 36.2B × 4.8 bits ÷ 8 = 21.7 GB
KV Cache + Overhead ≈ 0.6 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 69 GB (at full 524K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 4090 run Seed OSS 36B Base?
Yes, at Q4_K_M (22.3 GB) or lower. Higher quantizations like Q5_K_S (25.4 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.
What's the best quantization for Seed OSS 36B Base?
For Seed OSS 36B Base, Q4_K_M (22.3 GB) offers the best balance of quality and VRAM usage. Q5_K_S (25.4 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 10.5 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Seed OSS 36B Base on a Mac?
Yes — Mac Mini M2 Pro (16 GB) and 133 other Macs can run Seed OSS 36B Base, the smallest at IQ2_XXS. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 22.3 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio M1 Max (32 GB).
Can I run Seed OSS 36B Base locally?
Yes — Seed OSS 36B Base can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Seed OSS 36B Base?
At Q4_K_M, Seed OSS 36B Base can reach ~216 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~29 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 22.26 × 0.65 = ~234 tok/s
Estimated speed at Q4_K_M (22.3 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Seed OSS 36B Base?
At Q4_K_M, the download is about 21.69 GB. The full-precision BF16 version is 72.30 GB. The smallest option (IQ2_XXS) is 9.94 GB.
Which GPUs can run Seed OSS 36B Base?
7 consumer GPUs can run Seed OSS 36B Base at Q4_K_M. Top options include AMD Radeon RX 7900 XTX, NVIDIA GeForce RTX 3090.
Which devices can run Seed OSS 36B Base?
100 devices with unified memory can run Seed OSS 36B Base at Q4_K_M (22.3 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.