DeepSeek R1
DeepSeekLLMs
92.0
Performance
★ 4.6
Rating
480
Reviews
Text GenerationReasoningOpen WeightLarge
About
Open-weight reasoning model that matches OpenAI o1 on math and coding benchmarks through reinforcement learning and chain-of-thought.
Strengths
Matches or exceeds o1 on AIME 2024 (79.8%), MATH-500 (97.3%), and Codeforces. Open-weight with distilled smaller variants available. Uses extended thinking for complex problems. Revolutionary cost-performance for reasoning tasks.
Specifications
- Context window
- 128,000
- Parameters
- 671B
Pricing
- Input cost
- $0.55 / 1M
- Output cost
- $2.19 / 1M
$0.55 input (cache hit) / $2.19 output per 1M tokens. Open-weight for self-hosting.
Speed & Latency
- 25
- tokens/sec
- 1500ms
- time to first token
- 300 RPM
- rate limits
Available On
DeepSeek APIHuggingFaceTogether AIFireworks AI
Features
streamingreasoning tokenssystem messages
Performance Trend
Benchmark score trends over time for the top 5 benchmarks.
Loading history...
Benchmarks
Scores from various benchmark tests; higher is better.
| Test | Score | Percentile | Source |
|---|---|---|---|
| BigBench Hard | 89.0 | p99 | seed |
| GSM8K | 97.3 | p99 | seed |
| HumanEval | 90.5 | p98 | seed |
| MATH | 93.5 | p99 | seed |
