DeepSeek V3
A solid pick for self-hosting and full data control.
Overview
DeepSeek V3 is a revolution: 671B parameters in MoE, performance close to GPT-4o, but 30x lower price and MIT license (commercial use, self-hostable). Limits: French performance slightly behind, and for European enterprise, the official API is in China (prefer self-hosted deployment).
Skill profile
What do these scores mean?
PhD-level science questions (physics, chemistry, biology), with no tool access.
Resolving real GitHub issues under real conditions (SWE-bench Verified).
Competition-level math problems.
General knowledge across dozens of academic subjects.
Python code generation from specifications.
Strengths
- Excellent at code
- World-class on Arena
- Has a free tier
- Open-source and self-hostable
Limitations
- Average speed
Who is it for
- you want to control cost or self-host
- you have GDPR constraints
- you need very fast responses
Ideal use cases
- Ultra-low-cost coding
- Self-hosting
- Academic research
- Budget-tight apps
Access & availability
Key specifications
Estimate your monthly cost
Per-token API pricingFor the same usage
- Gemini 3 Flash$36+128%
- Gemini 3.1 Pro$145+812%
- GPT-5.5$363+2180%
Indicative estimate based on standard API rates (excluding caching, batch and volume discounts). Always check the official pricing before committing.
Privacy
Run it locally
| Quantization | Disk | RAM / VRAM | Typical hardware |
|---|---|---|---|
| Q4 · recommended | 390.4 GB | 431 GB | Server GPU infra (H100/A100…) |
| Q8 · balanced | 733 GB | 808 GB | Server GPU infra (H100/A100…) |
| FP16 · max quality | 1370 GB | 1509 GB | Server GPU infra (H100/A100…) |
Estimates for a moderate context. Long contexts need more RAM (KV cache).
Deploy
Copy-ready commands generated from this card. Adjust context length and GPU count to your hardware.
Easiest way to try it on a workstation. Install Ollama, then:
ollama run deepseek-v3Advanced data · for expertsArchitecture, modalities, detailed cost, full benchmarks▾
| Arena Elo | 1396 |
| MMLU | 87.5% |
| GPQA | 59.0% |
| HumanEval | 89.0% |
| SWE-Bench | 42.0% |
| MATH | 90.2% |