# DeepSeek R1

> Top reasoning model. 96% cheaper than o1.

Human version: https://kymaapi.com/models/deepseek-r1
Live JSON: `GET https://kymaapi.com/v1/models` (no auth required)

## Facts

- **Model ID**: `deepseek-r1` — pass this as `model` in the request body
- **Creator**: DeepSeek
- **Released**: 2025-01-20
- **Context window**: 64K tokens
- **Max output**: 33K tokens per response — a hard ceiling, not a default
- **Price**: $0.7309 in / $2.957 out per 1M
- **Capabilities**: tools, reasoning
- **Alias**: `reasoning` resolves to this model

## Call it

```bash
curl https://kymaapi.com/v1/chat/completions \
  -H "Authorization: Bearer $KYMA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "deepseek-r1", "messages": [{"role": "user", "content": "Hello"}]}'
```

## Positioning

DeepSeek R1 is a text-only reasoning model optimized for multi-step logic, mathematics, and structured analysis. Reach for it when you need explicit chain-of-thought capabilities at 96% lower cost than o1 and can tolerate slower generation speeds.

## About

DeepSeek R1 is a frontier reasoning model from DeepSeek that processes text inputs and returns text outputs. It is designed for tasks that require deep logical deduction, mathematical problem solving, and tool use, with native support for structured outputs.

On Kyma, the model runs behind automatic failover routing and returns exact request costs in the usage.cost field. It operates with a 64,000-token context window, a maximum output length of 32,768 tokens, and a median generation speed of 18 tokens per second. The reasoning alias resolves directly to this model.

This model does not support vision, prompt caching, or multimodal inputs. Because it runs on a slow speed tier, it is not suitable for latency-sensitive applications. Input and output tokens are billed per-token at the platform's listed rates.

## Use cases

- **Math & Logic Problems** — Solve complex mathematical and logical reasoning tasks with explicit step-by-step traces.
- **Document Analysis** — Extract structured insights and perform deep textual analysis across long documents.
- **Tool-Enabled Workflows** — Drive multi-step agent tasks that require explicit reasoning and function calling.

## Not ideal for

Avoid this model for real-time conversational interfaces, vision-based tasks, or applications that require prompt caching and low-latency responses.

## Pick something else when

- You need sub-second latency for chat → `qwen-3.7-plus`, `gemini-3-flash`
- You require vision or multimodal inputs → `qwen-3.7-max`, `gemini-3.5-flash`
- You need prompt caching for repeated prefixes → `qwen-3.6-plus`, `llama-3.3-70b`

## See also

- All models: https://kymaapi.com/models.md
- Pricing: https://kymaapi.com/pricing.md
- Other models by DeepSeek: https://kymaapi.com/models?q=DeepSeek