Overview
DeepSeek (Open Weights) is a large language model from DeepSeek. It supports a context window of 128K tokens and is designed for a broad range of reasoning, writing and coding tasks.
Exact capabilities, benchmark scores and pricing change frequently. Where current figures are not verified, this platform records them as "Unknown" rather than guessing.
Strengths
- General reasoning
- Following complex instructions
- Coding assistance
Weaknesses
- May hallucinate facts
- Knowledge cutoff limitations
Best use cases
Chat assistantsCoding copilotsContent generationData extraction
Benchmarks
Verified scores only — unknown values are labelled
| Benchmark | Category | Score |
|---|---|---|
| MMLU | Reasoning | Unknown |
| HumanEval | Coding | Unknown |
| GPQA | Reasoning | Unknown |
Frequently asked questions
Who makes DeepSeek (Open Weights)?
DeepSeek (Open Weights) is built by DeepSeek.
What is the context window of DeepSeek (Open Weights)?
DeepSeek (Open Weights) supports a context window of 128K tokens.
Does DeepSeek (Open Weights) support vision?
No, DeepSeek (Open Weights) is text-only.
How much does the DeepSeek (Open Weights) API cost?
Usage-based. See provider documentation for current per-token pricing.