DeepSeek R2
deepseek/deepseek-r2by DeepSeek·Text·164K context·Released 2026-04-30·MIT
Overview
DeepSeek's open reasoning model, the successor to R1, with long chain-of-thought and strong math, coding and planning. Competitive with the best hosted reasoners at a fraction of the cost, though reasoning tokens add to output spend. A favorite for hard problems where you still care about price.
Benchmarks
Standardized scores (0–100), normalized across the catalog.
Intelligence88
Coding86
Instruction following89
Providers
LLMRouteHub routes to the cheapest healthy provider by default. Showing the top 5 of 6 providers by routed volume.
| Provider | Input | Output | Throughput | Uptime |
|---|---|---|---|---|
| Hyperbolic | $0.55 | $2.40 | 58 tok/s | 99.4% |
| Parasail | $0.68 | $2.95 | 54 tok/s | 99.3% |
| DeepInfra | $0.80 | $3.50 | 50 tok/s | 99.1% |
| Together | $0.85 | $3.70 | 46 tok/s | 99% |
| Fireworks | $0.97 | $4.25 | 42 tok/s | 98.8% |
Use this model
curl https://llmroutehub.com/api/v1/chat/completions \
-H "Authorization: Bearer $LLMROUTEHUB_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek/deepseek-r2",
"messages": [{ "role": "user", "content": "Hello" }]
}'