DeepSeek R1
MIT Open SourceDeveloped by DeepSeek · MoE Reasoning Transformer · Modalities: Text · Released 2025-01-20
What are the token costs and operational benchmarks for DeepSeek R1?
Verified Capability Benchmarks
Composite Intelligence: 72/100Architectural & Production Profile
MIT-licensed open-weight reasoning powerhouse trained via large-scale reinforcement learning for mathematical and algorithmic synthesis.
Open-weight chain-of-thought reasoning, synthetic data distillation, and complex mathematical verification.
- Architecture
- MoE Reasoning Transformer
- Modalities
- Text
- Max Output Tokens
- 32,768 tokens
- Blended Rate (3:1)
- $0.960 / 1M
Verified Cloud Inference Endpoints Hosting DeepSeek R1
Compare latency, streaming speed, and token pricing across cloud API providers.
| Inference Provider | Role | Input / 1M | Output / 1M | Avg TTFT | Throughput |
|---|---|---|---|---|---|
| DeepSeek Platform → | Primary Host | $0.55 | $2.19 | 1450ms | 58 tok/s |
| Alibaba Cloud DashScope → | Partner Endpoint | $0.55 | $2.19 | 1330ms | 66.7 tok/s |
| Groq LPU Inference → | Partner Endpoint | $0.55 | $2.19 | 850ms | 150.8 tok/s |
| Cerebras CS-3 Wafer-Scale Inference → | Partner Endpoint | $0.55 | $2.19 | 700ms | 174 tok/s |
Interactive Monthly Token Economics & ROI Forecaster
Model production workloads across prompt, completion, and cached context tokens, or switch models on the fly.
Estimated Monthly Spend
$0.55/1M in · $2.19/1M out
Prompt Cache: $0.140/1M read · $0.55/1M write
$2.00/1M in · $10.00/1M out
Prompt Cache: $0.100/1M read · $0.50/1M write
Head-to-Head Comparisons Involving DeepSeek R1 (36)
DeepSeek R1 vs OpenAI o3-mini
DeepSeek R1 cuts input token costs by 50.0% compared to OpenAI o3-mini ($0.55 vs. $1.10 per 1M input tokens; $2.19 vs. $4.40/1M output). For autonomous software engineering, OpenAI o3-mini leads on SWE-bench Verified (49.3% vs. 49.2%), while OpenAI o3-mini edges ahead on GPQA Diamond scientific reasoning (79.7%). In real-time API production, OpenAI o3-mini achieves a faster 950ms TTFT latency (98.0 tokens/sec).
DeepSeek R1 vs Gemini 4 Argon
DeepSeek R1 is 72.5% cheaper for input tokens ($0.55 vs. $2.00 per 1M tokens) and $2.19 vs. $10.00 for output tokens (3.6x cost advantage). In terms of operational performance, Gemini 4 Argon delivers lower response latency at 850ms TTFT (600ms faster than DeepSeek R1) with 115 vs. 58 tok/s throughput. Gemini 4 Argon leads coding evaluations at 84.5% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 93.8% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Claude 4.5 Sonnet
DeepSeek R1 is 81.7% cheaper for input tokens ($0.55 vs. $3.00 per 1M tokens) and $2.19 vs. $15.00 for output tokens (5.5x cost advantage). In terms of operational performance, Claude 4.5 Sonnet delivers lower response latency at 380ms TTFT (1070ms faster than DeepSeek R1) with 84 vs. 58 tok/s throughput. Claude 4.5 Sonnet leads coding evaluations at 75.8% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 84.2% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Claude Opus 5.5
DeepSeek R1 is 86.2% cheaper for input tokens ($0.55 vs. $4.00 per 1M tokens) and $2.19 vs. $20.00 for output tokens (7.3x cost advantage). In terms of operational performance, Claude Opus 5.5 delivers lower response latency at 680ms TTFT (770ms faster than DeepSeek R1) with 95 vs. 58 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 93.2% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Gemini 3.8 Flash
Gemini 3.8 Flash is 54.5% cheaper for input tokens ($0.25 vs. $0.55 per 1M tokens) and $1.50 vs. $2.19 for output tokens (2.2x cost advantage). In terms of operational performance, Gemini 3.8 Flash delivers lower response latency at 105ms TTFT (1345ms faster than DeepSeek R1) with 210 vs. 58 tok/s throughput. Gemini 3.8 Flash leads coding evaluations at 68.5% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 79.2% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs GPT-5
DeepSeek R1 is 78.0% cheaper for input tokens ($0.55 vs. $2.50 per 1M tokens) and $2.19 vs. $10.00 for output tokens (4.5x cost advantage). In terms of operational performance, GPT-5 delivers lower response latency at 340ms TTFT (1110ms faster than DeepSeek R1) with 105 vs. 58 tok/s throughput. GPT-5 leads coding evaluations at 74.9% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 85.4% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs GPT-6 Astra
DeepSeek R1 is 94.5% cheaper for input tokens ($0.55 vs. $10.00 per 1M tokens) and $2.19 vs. $50.00 for output tokens (18.2x cost advantage). In terms of operational performance, GPT-6 Astra delivers lower response latency at 740ms TTFT (710ms faster than DeepSeek R1) with 74 vs. 58 tok/s throughput. GPT-6 Astra leads coding evaluations at 75.7% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 96.0% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs DeepSeek V4 Pro
DeepSeek R1 is 58.3% cheaper for input tokens ($0.55 vs. $1.32 per 1M tokens) and $2.19 vs. $3.96 for output tokens (2.4x cost advantage). In terms of operational performance, DeepSeek V4 Pro delivers lower response latency at 520ms TTFT (930ms faster than DeepSeek R1) with 78 vs. 58 tok/s throughput. DeepSeek V4 Pro leads coding evaluations at 80.6% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 90.1% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs DeepSeek V4 Flash Vision (Exp)
DeepSeek V4 Flash Vision (Exp) outperforms DeepSeek R1 across capability, speed, and cost. V4 cuts input pricing by 60.0% ($0.22 vs. $0.55) and output to $0.65 vs. $2.19 per 1M tokens. It leads SWE-bench Verified (77.4% vs. 49.2%) and GPQA Diamond (91.3% vs. 71.5%), achieving 310ms TTFT (1140ms faster) at 165 vs. 58 tok/s.
DeepSeek R1 vs OpenAI o3
DeepSeek R1 is 72.5% cheaper for input tokens ($0.55 vs. $2.00 per 1M tokens) and $2.19 vs. $8.00 for output tokens (3.6x cost advantage). In terms of operational performance, OpenAI o3 delivers lower response latency at 1150ms TTFT (300ms faster than DeepSeek R1) with 82 vs. 58 tok/s throughput. OpenAI o3 leads coding evaluations at 71.7% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 87.7% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Claude 4 Opus
DeepSeek R1 is 96.3% cheaper for input tokens ($0.55 vs. $15.00 per 1M tokens) and $2.19 vs. $75.00 for output tokens (27.3x cost advantage). In terms of operational performance, Claude 4 Opus delivers lower response latency at 920ms TTFT (530ms faster than DeepSeek R1) with 46 vs. 58 tok/s throughput. Claude 4 Opus leads coding evaluations at 77.2% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 87.1% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs o1
DeepSeek R1 cuts input token costs by 96.3% compared to o1 ($0.55 vs. $15.00 per 1M input tokens; $2.19 vs. $60.00/1M output). For autonomous software engineering, DeepSeek R1 leads on SWE-bench Verified (49.2% vs. 48.9%), while o1 edges ahead on GPQA Diamond scientific reasoning (78.0%). In real-time API production, DeepSeek R1 achieves a faster 1450ms TTFT latency (58.0 tokens/sec).
DeepSeek R1 vs Claude 3.7 Sonnet
DeepSeek R1 is 81.7% cheaper for input tokens ($0.55 vs. $3.00 per 1M tokens) and $2.19 vs. $15.00 for output tokens (5.5x cost advantage). In terms of operational performance, Claude 3.7 Sonnet delivers lower response latency at 540ms TTFT (910ms faster than DeepSeek R1) with 72 vs. 58 tok/s throughput. Claude 3.7 Sonnet leads coding evaluations at 70.3% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 78.5% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Gemini 3.1 Pro
DeepSeek R1 is 56.0% cheaper for input tokens ($0.55 vs. $1.25 per 1M tokens) and $2.19 vs. $10.00 for output tokens (2.3x cost advantage). In terms of operational performance, Gemini 3.1 Pro delivers lower response latency at 420ms TTFT (1030ms faster than DeepSeek R1) with 98 vs. 58 tok/s throughput. Gemini 3.1 Pro leads coding evaluations at 71.8% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 82.4% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Grok 4
DeepSeek R1 is 78.0% cheaper for input tokens ($0.55 vs. $2.50 per 1M tokens) and $2.19 vs. $12.50 for output tokens (4.5x cost advantage). In terms of operational performance, Grok 4 delivers lower response latency at 480ms TTFT (970ms faster than DeepSeek R1) with 95 vs. 58 tok/s throughput. Grok 4 leads coding evaluations at 73.2% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 86.2% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Llama 4 Maverick
Llama 4 Maverick is 63.6% cheaper for input tokens ($0.20 vs. $0.55 per 1M tokens) and $0.60 vs. $2.19 for output tokens (2.8x cost advantage). In terms of operational performance, Llama 4 Maverick delivers lower response latency at 135ms TTFT (1315ms faster than DeepSeek R1) with 175 vs. 58 tok/s throughput. Llama 4 Maverick leads coding evaluations at 62.4% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 73.8% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Qwen 3.8 Flash Next
Qwen 3.8 Flash Next is 70.9% cheaper for input tokens ($0.16 vs. $0.55 per 1M tokens) and $0.47 vs. $2.19 for output tokens (3.4x cost advantage). In terms of operational performance, Qwen 3.8 Flash Next delivers lower response latency at 280ms TTFT (1170ms faster than DeepSeek R1) with 185 vs. 58 tok/s throughput. Qwen 3.8 Flash Next leads coding evaluations at 81.0% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 91.7% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Gemini 3.5 Flash-Lite
Gemini 3.5 Flash-Lite is 86.4% cheaper for input tokens ($0.07 vs. $0.55 per 1M tokens) and $0.30 vs. $2.19 for output tokens (7.3x cost advantage). In terms of operational performance, Gemini 3.5 Flash-Lite delivers lower response latency at 78ms TTFT (1372ms faster than DeepSeek R1) with 265 vs. 58 tok/s throughput. Gemini 3.5 Flash-Lite leads coding evaluations at 55.4% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 68.4% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs OpenAI o4-mini
DeepSeek R1 is 50.0% cheaper for input tokens ($0.55 vs. $1.10 per 1M tokens) and $2.19 vs. $4.40 for output tokens (2.0x cost advantage). In terms of operational performance, OpenAI o4-mini delivers lower response latency at 680ms TTFT (770ms faster than DeepSeek R1) with 125 vs. 58 tok/s throughput. OpenAI o4-mini leads coding evaluations at 68.1% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 81.4% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs DeepSeek V3
DeepSeek V3 is 74.5% cheaper for input tokens ($0.14 vs. $0.55 per 1M tokens) and $0.28 vs. $2.19 for output tokens (3.9x cost advantage). In terms of operational performance, DeepSeek V3 delivers lower response latency at 410ms TTFT (1040ms faster than DeepSeek R1) with 78 vs. 58 tok/s throughput. DeepSeek R1 leads coding evaluations at 49.2% SWE-bench Verified vs. 42.0% for DeepSeek V3 and 71.5% vs. 59.1% on GPQA Diamond.
DeepSeek R1 vs Gemma 4 31B IT
Gemma 4 31B IT is 78.2% cheaper for input tokens ($0.12 vs. $0.55 per 1M tokens) and $0.36 vs. $2.19 for output tokens (4.6x cost advantage). In terms of operational performance, Gemma 4 31B IT delivers lower response latency at 115ms TTFT (1335ms faster than DeepSeek R1) with 155 vs. 58 tok/s throughput. Gemma 4 31B IT leads coding evaluations at 59.2% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 71.2% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Claude 3.5 Haiku
DeepSeek R1 is 31.2% cheaper for input tokens ($0.55 vs. $0.80 per 1M tokens) and $2.19 vs. $4.00 for output tokens (1.5x cost advantage). In terms of operational performance, Claude 3.5 Haiku delivers lower response latency at 190ms TTFT (1260ms faster than DeepSeek R1) with 130 vs. 58 tok/s throughput. DeepSeek R1 leads coding evaluations at 49.2% SWE-bench Verified vs. 40.6% for Claude 3.5 Haiku and 71.5% vs. 41.6% on GPQA Diamond.
DeepSeek R1 vs Claude 3.5 Sonnet
DeepSeek R1 is 81.7% cheaper for input tokens ($0.55 vs. $3.00 per 1M tokens) and $2.19 vs. $15.00 for output tokens (5.5x cost advantage). In terms of operational performance, Claude 3.5 Sonnet delivers lower response latency at 310ms TTFT (1140ms faster than DeepSeek R1) with 80 vs. 58 tok/s throughput. DeepSeek R1 leads coding evaluations at 49.2% SWE-bench Verified vs. 49.0% for Claude 3.5 Sonnet and 71.5% vs. 59.4% on GPQA Diamond.
DeepSeek R1 vs Gemini 2.0 Flash
Gemini 2.0 Flash is 81.8% cheaper for input tokens ($0.10 vs. $0.55 per 1M tokens) and $0.40 vs. $2.19 for output tokens (5.5x cost advantage). In terms of operational performance, Gemini 2.0 Flash delivers lower response latency at 280ms TTFT (1170ms faster than DeepSeek R1) with 185 vs. 58 tok/s throughput. Gemini 2.0 Flash leads coding evaluations at 51.8% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 63.6% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Gemini 2.5 Flash
Gemini 2.5 Flash is 45.5% cheaper for input tokens ($0.30 vs. $0.55 per 1M tokens) and $2.50 vs. $2.19 for output tokens (1.8x cost advantage). In terms of operational performance, Gemini 2.5 Flash delivers lower response latency at 110ms TTFT (1340ms faster than DeepSeek R1) with 180 vs. 58 tok/s throughput. Gemini 2.5 Flash leads coding evaluations at 58.0% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 70.5% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Gemini 2.5 Pro
DeepSeek R1 is 56.0% cheaper for input tokens ($0.55 vs. $1.25 per 1M tokens) and $2.19 vs. $10.00 for output tokens (2.3x cost advantage). In terms of operational performance, Gemini 2.5 Pro delivers lower response latency at 700ms TTFT (750ms faster than DeepSeek R1) with 80 vs. 58 tok/s throughput. Gemini 2.5 Pro leads coding evaluations at 63.8% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 74.0% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs GLM-5.3
GLM-5.3 is 36.4% cheaper for input tokens ($0.35 vs. $0.55 per 1M tokens) and $1.00 vs. $2.19 for output tokens (1.6x cost advantage). In terms of operational performance, GLM-5.3 delivers lower response latency at 160ms TTFT (1290ms faster than DeepSeek R1) with 120 vs. 58 tok/s throughput. GLM-5.3 leads coding evaluations at 56.0% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 67.0% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs GLM-5.3-Flash
GLM-5.3-Flash is 78.2% cheaper for input tokens ($0.12 vs. $0.55 per 1M tokens) and $0.40 vs. $2.19 for output tokens (4.6x cost advantage). In terms of operational performance, GLM-5.3-Flash delivers lower response latency at 90ms TTFT (1360ms faster than DeepSeek R1) with 175 vs. 58 tok/s throughput. GLM-5.3-Flash leads coding evaluations at 52.8% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 63.5% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs GPT-4o
DeepSeek R1 is 78.0% cheaper for input tokens ($0.55 vs. $2.50 per 1M tokens) and $2.19 vs. $10.00 for output tokens (4.5x cost advantage). In terms of operational performance, GPT-4o delivers lower response latency at 280ms TTFT (1170ms faster than DeepSeek R1) with 100 vs. 58 tok/s throughput. DeepSeek R1 leads coding evaluations at 49.2% SWE-bench Verified vs. 48.0% for GPT-4o and 71.5% vs. 53.6% on GPQA Diamond.
DeepSeek R1 vs GPT-4o Mini
GPT-4o Mini is 72.7% cheaper for input tokens ($0.15 vs. $0.55 per 1M tokens) and $0.60 vs. $2.19 for output tokens (3.7x cost advantage). In terms of operational performance, GPT-4o Mini delivers lower response latency at 220ms TTFT (1230ms faster than DeepSeek R1) with 140 vs. 58 tok/s throughput. DeepSeek R1 leads coding evaluations at 49.2% SWE-bench Verified vs. 41.0% for GPT-4o Mini and 71.5% vs. 40.2% on GPQA Diamond.
DeepSeek R1 vs GPT-6.1 Sol
DeepSeek R1 is 72.5% cheaper for input tokens ($0.55 vs. $2.00 per 1M tokens) and $2.19 vs. $10.00 for output tokens (3.6x cost advantage). In terms of operational performance, GPT-6.1 Sol delivers lower response latency at 520ms TTFT (930ms faster than DeepSeek R1) with 135 vs. 58 tok/s throughput. GPT-6.1 Sol leads coding evaluations at 86.4% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 92.4% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Grok 3
DeepSeek R1 is 81.7% cheaper for input tokens ($0.55 vs. $3.00 per 1M tokens) and $2.19 vs. $15.00 for output tokens (5.5x cost advantage). In terms of operational performance, Grok 3 delivers lower response latency at 620ms TTFT (830ms faster than DeepSeek R1) with 88 vs. 58 tok/s throughput. Grok 3 leads coding evaluations at 65.8% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 84.6% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Grok 3 Mini
Grok 3 Mini is 45.5% cheaper for input tokens ($0.30 vs. $0.55 per 1M tokens) and $1.20 vs. $2.19 for output tokens (1.8x cost advantage). In terms of operational performance, Grok 3 Mini delivers lower response latency at 150ms TTFT (1300ms faster than DeepSeek R1) with 160 vs. 58 tok/s throughput. Grok 3 Mini leads coding evaluations at 54.0% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 68.0% vs. 71.5% on GPQA Diamond.
DeepSeek R1 vs Llama 3.3 70B Instruct
Llama 3.3 70B Instruct is 78.2% cheaper for input tokens ($0.12 vs. $0.55 per 1M tokens) and $0.30 vs. $2.19 for output tokens (4.6x cost advantage). In terms of operational performance, Llama 3.3 70B Instruct delivers lower response latency at 145ms TTFT (1305ms faster than DeepSeek R1) with 165 vs. 58 tok/s throughput. DeepSeek R1 leads coding evaluations at 49.2% SWE-bench Verified vs. 42.5% for Llama 3.3 70B Instruct and 71.5% vs. 55.6% on GPQA Diamond.
DeepSeek R1 vs Mistral Large 2 (2411)
DeepSeek R1 is 72.5% cheaper for input tokens ($0.55 vs. $2.00 per 1M tokens) and $2.19 vs. $6.00 for output tokens (3.6x cost advantage). In terms of operational performance, Mistral Large 2 (2411) delivers lower response latency at 380ms TTFT (1070ms faster than DeepSeek R1) with 75 vs. 58 tok/s throughput. DeepSeek R1 leads coding evaluations at 49.2% SWE-bench Verified vs. 45.2% for Mistral Large 2 (2411) and 71.5% vs. 54.2% on GPQA Diamond.
DeepSeek R1 vs Qwen 3.8 27B
Qwen 3.8 27B is 63.6% cheaper for input tokens ($0.20 vs. $0.55 per 1M tokens) and $0.60 vs. $2.19 for output tokens (2.8x cost advantage). In terms of operational performance, Qwen 3.8 27B delivers lower response latency at 120ms TTFT (1330ms faster than DeepSeek R1) with 140 vs. 58 tok/s throughput. Qwen 3.8 27B leads coding evaluations at 58.4% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 68.0% vs. 71.5% on GPQA Diamond.
Frequently Asked Questions & Query Fan-Out
How much does DeepSeek R1 cost per 1M tokens including prompt caching?
DeepSeek R1 costs $0.55 per million input tokens and $2.19 per million output tokens. Cached input reads cost $0.140 per 1M tokens, reducing repeated context costs by 75%.
What are DeepSeek R1's coding and reasoning benchmark scores?
DeepSeek R1 achieves 49.2% on SWE-bench Verified, 92.8% on HumanEval coding, 71.5% on GPQA Diamond graduate-level science, 84% on MMLU-Pro, and 97.3% on MATH-500, earning an overall Inference Value Ratio (IVR) of 64.8/100.
What is the context window, TTFT latency, and token throughput of DeepSeek R1?
DeepSeek R1 supports a 131,072 token context window (32,768 max output tokens) and operates at 1450ms average Time to First Token (TTFT) with 58 tokens/sec output streaming throughput.
What are the primary production use cases for DeepSeek R1?
Open-weight chain-of-thought reasoning, synthetic data distillation, and complex mathematical verification.