YOUR AI, WITH YOUR CHOSEN MODEL

GLM-5.3-Flash

A large model for development and tool use. Placement and memory headroom will be tested on the selected server.

We’ll prepare the hardware, model and a convenient interface5,710,000–6,400,000 RUBPreliminary complete-system price Atlas 336View complete systems

COMPLETE-SYSTEM CONFIGURATION AND PRICE

Complete systems for this model.

Start with a complete configuration. Before ordering, we clarify tasks and workload, confirm compatibility and test speed.

For DeepSeek and large modelsBuilt to order

Atlas 336

Four professional GPUs for development, agents and complex tasks.

Preliminary complete-system price5,710,000–6,400,000 RUBHardware, assembly and initial setup

Graphics cards: 4 × RTX PRO 6000D 84 GB
Memory / storage: 256 GB DDR5 ECC RDIMM / 2 + 4 TB

GLM-5.3-FlashFP8

DeepSeek V4.1 Flash profile
One server with four GPUs

Choose Atlas 336

Full configuration and terms

Processor / motherboard: Threadripper PRO 9965WX · 24 cores / ASUS WRX90
Graphics cards: 4 × RTX PRO 6000D 84 GB
System memory: 256 GB DDR5 ECC RDIMM · 8 × 32 GB
Drives: 2 TB for system + 4 TB for models
Power and cooling: At least 3000 W usable power · server airflow or factory liquid cooling
Model placement: For DeepSeek: four GPUs in one group, with part of the model in RAM

We test speed and your workloads before handover. Shipping, import and complex integrations are charged separately.

What is included in the price and how is it calculated?

Estimate dated 12 October 2026: the complete hardware system plus allowances for assembly, initial setup of the selected model and testing. Handover in China. Shipping, import, ongoing support and complex integrations are calculated separately.

This preliminary range uses public component and base-platform prices, budget allowances and a 10–15% contingency. Before ordering, the quotation will confirm final price, compatibility, availability, lead time and warranty. This range is not a system integrator’s quotation.

View calculations and sources

Tested on your tasks before handover.Time to first response, speed and concurrent operation are part of the agreed test. Headcount alone does not determine workload.

MANUFACTURER ALTERNATIVES

Configurations from established manufacturers.

These platforms are candidates for testing the chosen version. The system, price and operating profile will be confirmed before ordering.

LenovoTower / rack-mountable

For DeepSeek and agents

ThinkStation PX

A powerful factory workstation for programming and shared local AI access.

4 × RTX PRO 6000 Blackwell Max-Q 96 GB512 GB ECC RDIMM · 16 × 32 GB384 GB combined GPU memory · model placement subject to testing

Our proposed configuration for a factory enquiry

Complete systemPrice on requestWe’ll confirm price and supply availability

MODELS TO TESTDeepSeek-V4.1-Flash · FP4/FP8 GLM-5.3-Flash · FP8 MiniMax-M2.7 · FP8

Request price and lead time

Configuration and verification

CPU: 2 × Xeon Gold 6530 · 32 cores each
RAM: 512 GB ECC RDIMM · 16 × 32 GB
Storage: NVMe M.2: 2 TB + 4 TB
Power: Factory PX power system, variant matched to the selected BOM

Officially supports up to four Max-Q 300 W GPUs. CPU, RAM and SSD are proposed. DeepSeek: about 188.83 GiB of Engram in RAM; engine and speed require testing. GPU support does not mean the model has been tested.

GPU support is published by the manufacturer. The selected model has not yet been tested.

Official specification

Lenovo3U rack

Server for an AI service

ThinkSystem SR675 V3

Four server GPUs for a shared interface, API and the selected large model.

4 × RTX PRO 6000 Blackwell Server Edition 96 GB768 GB ECC RDIMM · 24 × 32 GB384 GB combined GPU memory · model placement subject to testing

Our proposed configuration for a factory enquiry

Complete systemPrice on requestWe’ll confirm price and supply availability

MODELS TO TESTDeepSeek-V4.1-Flash · FP4/FP8 GLM-5.3-Flash · FP8 MiniMax-M2.7 · FP8

Request price and lead time

Configuration and verification

CPU: 2 × AMD EPYC 9535 · 64 cores each
RAM: 768 GB ECC RDIMM · 24 × 32 GB
Storage: 2 × 960 GB M.2 + 2 × 3,84 TB NVMe
Power: 4 × 2600 W Titanium; usable power depends on redundancy

CBK8 requires EPYC 9005 and Dual Rotor High Performance Fan C5WW. The specific CPU/RAM/SSD choices are our configuration enquiry, not a fixed factory SKU.

GPU support is published by the manufacturer. The selected model has not yet been tested.

Official specification

Lenovo3U rack

For large original models

ThinkSystem SR675i V3 · CF3F

A fixed factory configuration for demanding AI services and several work scenarios.

8 × RTX PRO 6000 Blackwell Server Edition 96 GB1,5 TB ECC RDIMM · 24 × 64 GB768 GB combined GPU memory · model placement subject to testing

Fixed manufacturer configuration

Complete systemPrice on requestWe’ll confirm price and supply availability

MODELS TO TESTGLM-5.3 · FP8 MiniMax-M3-MXFP8 · MXFP8 DeepSeek-V4.1-Flash · FP4/FP8

Request price and lead time

Configuration and verification

CPU: 2 × AMD EPYC 9535 · 64 cores each
RAM: 1,5 TB ECC RDIMM · 24 × 64 GB
Storage: 2 × 960 GB M.2 + 2 × 3,84 TB E3.S NVMe
Power: 4 × 2600 W Titanium · 230 V

Official CF3F BOM: 4 × BlueField-3 B3140H 400G and 1 × B3220 2 × 200G, CBK8 GPUs. Main components are fixed in CTO. 768 GB is the sum of GPU memory, not a single memory bank. GLM files are about 703.74 GiB: runtime headroom and engine must be checked separately.

GPU support is published by the manufacturer. The selected model has not yet been tested.

Official specification

Manufacturer catalogue

Technical model detailsFormat, memory, exact version and sizing conditions

Approximately 305.79 GiB of weight files. A 336 GB platform passes the preliminary capacity comparison, but leaves little room for cache and buffers. Actual placement must be tested.

WEIGHT FILES305.79 GiB

FORMATOfficial FP8

NATIVE CONTEXTNot confirmed in the review

Example technical selection: 4 simultaneous requests, 32K context, BF16 KV. These are example conditions, not your requirements. Preliminary calculation for the exact checkpoint. Speed, actual memory use and the selected workload have not yet been measured. The native context limit is not confirmed here; the selected length needs testing.

Size is the sum of safetensors files, not measured memory use during inference. Context, buffers, loading and execution need additional resources.

Repository: zai-org/GLM-5.3-Flash
Revision: eb9eb208eb0d988989d07a6a12d0fdeb5f52574a
File size: 328,337,455,672 bytes
Verified: 11 October 2026

Official model version Configure model and workload

Need help choosing?

Find a system for my task