🇨🇳 China💻 CodeReleased February 2025
DeepSeek Coder V3
DeepSeek
Open-sourceFreeSelf-host
Verdict
A solid pick for self-hosting and full data control.
Overview
DeepSeek Coder V3 is the reference model for low-cost code. It beats Claude Sonnet on HumanEval at 50x lower price, supporting 300+ languages. MIT license, 100% free.
Skill profile
Arena Elo1280
Verified benchmarks
SWE-Bench49%
HumanEval90.2%
What do these scores mean?
SWE-Bench49%fair
Resolving real GitHub issues under real conditions (SWE-bench Verified).
HumanEval90.2%world-class
Python code generation from specifications.
Strengths
- Excellent at code
- Has a free tier
- Open-source and self-hostable
- Competitive input pricing
Limitations
- No native GDPR guarantee
- Light safety filters
Who is it for
A good fit if…
- you build with a coding agent
- you want to control cost or self-host
Skip it if…
- you handle sensitive EU data
- you want strict guardrails
Ideal use cases
- Low-cost code autocompletion
- Code review
- Boilerplate generation
- 300+ languages
Access & availability
Paid API per tokenFree tier availableSelf-hostable (open weights)
Key specifications
Context
128K
Input price
$0.14 $/M
Output price
$0.28 $/M
Speed
80 tok/s
Price not auditedBenchmarks not audited
Privacy
Not GDPR-compliantHosting : CN/selfAnonymizable data
Advanced data · for expertsArchitecture, modalities, detailed cost, full benchmarks▾
Advanced data · for experts
Architecture, modalities, detailed cost, full benchmarks
Architecture
MoE
Size
236B (21B actifs MoE)
Cutoff date
Nov 2024
Inputs
text
Outputs
code
License
mit
Hosting
CN/self
Hallucination score
3/5
Estimated cost (API)
Typical exchange (~3k in / 1k out)
≈ $0.0007
1M in + 1M out
≈ $0.42
All benchmarks
| Arena Elo | 1280 |
| MMLU | — |
| GPQA | — |
| HumanEval | 90.2% |
| SWE-Bench | 49.0% |
| MATH | — |