Model · DeepSeek
DeepSeek
Open-weights models that punch above their price, especially on reasoning.
DeepSeek earned its reputation by delivering strong reasoning at a fraction of the cost of the big closed models. The V4 line adds very long context, and R1 is built specifically for step-by-step reasoning.
The lineup
| Model | Best for | Plan |
|---|---|---|
| DeepSeek V4 Flash | Fast reasoning with 1M-token context. Included free. | Free |
| DeepSeek V4 Pro | The top reasoning tier with long context. | Paid |
| DeepSeek R1 | Explicit step-by-step reasoning. | Paid |
| DeepSeek V3.2 / V3 Chat | Stable general chat variants. | Paid |
What it is good for
- Strong reasoning without a premium price
- Very long inputs, thanks to the V4 million-token context
- A high-value free option in the mix
- Adding a different reasoning style to a comparison
In a weave
DeepSeek V4 Flash is one of the free models, and it brings real reasoning to free-plan weaves, not just a lightweight filler answer.
On paid plans, V4 Pro and R1 are good additions when you want a reasoning-heavy model alongside the usual frontier names.
DeepSeek V4 Flash is included free. The other DeepSeek variants are on the paid plans.
Common questions
Is DeepSeek good for reasoning?
Yes, that is its strength. R1 is built for explicit step-by-step reasoning, and the V4 line reasons well while handling very long inputs.
Is DeepSeek free on LLMWeave?
DeepSeek V4 Flash is part of the free model set. The Pro, R1, and V3 variants are on the paid plans.
More models
Claude
Anthropic’s models. Careful reasoning, strong writing, and steady instruction-following.
OpenAIGPT
OpenAI’s models. Versatile all-rounders with strong general reasoning.
GoogleGemini
Google’s models. Huge context windows and fast, cheap Flash tiers.
AlibabaQwen
Alibaba’s models. Strong multilingual range, plus solid coding and math.
MetaLlama
Meta’s open-weights models. Efficient, compact, and widely adopted.
MistralMistral
Efficient European models, a good fit for synthesis and judging.
xAIGrok
xAI’s models. A distinct voice and conversational style.
Try DeepSeek in a weave
Run it next to other models on your own task. Free to start, no card.
Get started