Skip to content
English

Switch theme

// ByteDance Seed · Text LLM

Seed-Coder-8B-Instruct: VRAM & local deployment

8.25B parameters. About 7.1 GB at 4-bit and 19.7 GB at BF16 with one 8K-token request.

released 2025-04-27licence: Commercial OKmemory figures are estimates

01Specs

PublisherByteDance Seed · ByteDance-Seed
Family / seriesByteDance Seed · Seed-Coder
TypeText LLM (text-generation)
Parameters8.25B Source: counted from official safetensors weights
Context length32K tokens (32,768)max_position_embeddings in config.json
Architecture32 layers · GQA
Released weightsBF16 · 16.5 GBdownload
LicenceMIT · Commercial OKPermissive licence (Apache-2.0 / MIT / CC-BY / OpenMDW): commercial use allowed.
Released2025-04-27 (Hugging Face repo created)

02How much VRAM it needs (estimate)

Total = weights + 10% + 0.5GB overhead + KV cache for one request (FP16 KV). More concurrent users multiply the KV column.

PrecisionWeightsTotal @ 8KTotal @ 32KSmallest GPUMac
BF16 / FP16 (original)16.5 GB 19.7 GB22.9 GB RTX 4090 24GBMac 36GB
FP8 / INT8 / Q8_08.8 GB 11.2 GB14.4 GB 16GB GPUMac 24GB
4-bit (Q4_K_M / AWQ / GPTQ-Int4)5.0 GB 7.1 GB10.3 GB 8GB GPUMac 16GB

KV cache per request: 8K 1.1 GB · 32K 4.3 GB

Bits per weight: BF16 16, FP8 8.5, 4-bit 4.85. “Fits” = total ≤ 90% of memory; Mac GPU can use ~2/3 of unified memory (≤36GB) or ~3/4 (larger). For MoE, all experts must be in memory, so size by total parameters. Method →

03How to run it

Only runtimes the publisher or the runtime’s official library names for this model.

Which inference engine to choose →

04Official repositories & quantised builds

2 official repo(s) from ByteDance-Seed.

05Other ByteDance Seed models

All 9 →
ModelParams4-bit totalReleased
Seed-OSS-36B-Instruct36.2B26.8 GB2025-08-20
Seed-X-Instruct-7B7.51B6.6 GB2025-07-16
Seed-X-PPO-7B7.51B6.6 GB2025-07-16
SeedVR-3B3B2.5 GB2025-06-11
SeedVR2-7B7B5.2 GB2025-06-11
SeedVR2-3B3B2.5 GB2025-06-10
BAGEL-7B-MoT14.7B10.8 GB2025-05-19
UI-TARS-1.5-7B8.29B6.5 GB2025-04-16

06Sources

Data checked 2025-06-06. Parameters, context and licence are copied from the official repository; memory figures are our estimates.

Data updated: Self-host open models(UTC+8)