The CORSAIR FlexPrime V20C is a desktop workstation for developing and running local AI. You can use it to explain code, compare spreadsheets, or build an assistant that answers questions about your documents.
This configuration pairs an Intel Core Ultra 7 270K Plus processor with Intel Arc Pro B70 graphics. Here’s a look at the hardware, two model options, and an Excel comparison in Hermes Desktop.
The specifications below apply to SKU CSW-9040004-NA.
| Component | Specification |
|---|---|
| Processor | Intel Core Ultra 7 270K Plus |
| Graphics | Intel Arc Pro B70 |
| System memory | 32GB DDR5, installed as 2 × 16GB |
| Storage | 1TB M.2 NVMe SSD |
| Operating system | Ubuntu 24.04 LTS |
| Motherboard platform | Intel Z890 |
| Power supply | 850W |
The Core Ultra 7 270K Plus has 24 cores and 24 threads. The Arc Pro B70 has 32GB of dedicated GDDR6 graphics memory.
AI models running on the B70 use its dedicated graphics memory. Ubuntu and your other apps use the workstation’s separate 32GB of DDR5 RAM.
Longer documents, more conversation history, and several requests running at once can all increase memory use.
Community recipes describe running compact versions of Qwen3.8-27B and NVIDIA Nemotron 3.5 Lightning 30B-A3B on B70 hardware. These versions are quantized, meaning they store the model more compactly to reduce memory use.
| Model | What you could use it for | Download option |
|---|---|---|
| Qwen3.8-27B | Explain code or compare documents. | Q4_K_M, about 19GB |
| NVIDIA Nemotron 3.5 Lightning 30B-A3B | Build a text-based assistant for chat, coding help, or questions about documents. | Community Q4_K_S version, about 23.2GB |
Both downloads use GGUF, a model file format supported by llama.cpp, the software that loads and runs the model.
Choose the Q4_K_M model file. The repository total includes other model versions and companion files.
Comparing two versions of a budget means finding changed prices and quantities, added items, and costs that have disappeared. Hermes Desktop can read Excel attachments and summarize those differences in a conversation.
It can connect to a local model server. The server runs the model, while Hermes supplies the conversation and tools for working with your files. Here’s how to approach that setup with Qwen.
Follow the Qwen3.8 B70 guide to build the server. Under qwen38-b70 in docker-compose.yml, use these settings for one B70, one request at a time, and a 65,536-token context. The paths point to the model folder and startup script in the cloned recipe.
volumes:
- /dev/dri/by-path:/dev/dri/by-path:ro
- ./models:/models
- ./entrypoint.sh:/entrypoint.sh:ro
environment:
- MODELS_DIR=/models
- GPU_COUNT=1
- CTX_SIZE_OVERRIDE=65536
- PARALLEL=1
- ENABLE_MTP=0
ports:
- "127.0.0.1:8010:8010"The recipe downloads its matching Qwen model when it first starts. Add --jinja to the llama-server command in entrypoint.sh to use Qwen’s chat template for tool calls.
In Hermes Desktop, open Settings > Providers > Custom Endpoints and choose Chat Completions. Enter http://127.0.0.1:8010/v1 as the server address. Use port 8010 for this setup; the screenshot shows an example using 8080.
Enter the model identifier shown by your local server. Hermes connects to the model running on your workstation through this address. If your server doesn’t require authentication, leave the API key field blank.
Select Test to check the connection, then choose Qwen for your conversation. Attach both sample Excel budgets and use the prompt below to compare them.
Open Settings > Model > Custom endpoint > Set up Custom endpoint > Local / custom endpoint.
Download event-budget-v1.xlsx and event-budget-v2.xlsx, then attach both files to a Hermes conversation. These sample workbooks contain fictional event budgets, with changes to quantities, prices, and items between versions. Try this prompt:
Compare event-budget-v1.xlsx with event-budget-v2.xlsx, matching rows by Item ID. List added and removed items, changes to quantities or unit costs, and each change’s effect on the budget. Present the changes as a Markdown table with columns Item, Change, and Budget difference. Give the total for each version and the overall difference in USD.
Hermes matches the items by ID, identifies the changes, and calculates the difference between the two budgets:
The revised budget totals $1,707, up $234 from the original $1,473. Display hire and catering account for the largest increases. Room rental and delivery stay the same.
The V20C uses a 360mm liquid CPU cooler, five 120mm fans, and an air-cooled GPU. It weighs 14kg and comes with a three-year warranty and lifetime technical support.
If you need more GPU memory, storage, or processing power, our CORSAIR PRO range includes compact and full-tower workstations as well as rackmount GPU servers.
The CORSAIR PRO configurator lets you start with a preset or choose individual components, including the processor, graphics, memory, storage, and operating system. The configured price updates as you make selections. You can then view a build summary and request a quote for your chosen configuration.
PRODUCTS IN ARTICLE
Stay up to date with CORSAIR. Get our latest News, Guides, and Product Updates in your Google feeds.
Add CORSAIR as a preferred sourceThanks for helping us improve this article.
JOIN OUR OFFICIAL CORSAIR COMMUNITIES
Join our official CORSAIR Communities! Whether you're new or old to PC Building, have questions about our products, or want to chat about the latest PC, tech, and gaming trends, our community is the place for you.