DeepSeek R2
DeepSeek · released 2026-02-15
Quality (AA Index)86
Input price$0.55
Output price$2.20
Context128K
Throughput110 tok/s
P50 latency1.2s
Side-by-side pricing, capabilities, real-world cost across common scenarios, and our editorial pick.
DeepSeek · released 2026-02-15
Google · released 2025-09-01
链接通过我们的 OpenRouter 推荐链接在新标签页打开。 查看推广披露
| Metric | DeepSeek R2 | Gemini 2.5 Pro | Verdict |
|---|---|---|---|
Input price /百万 tokens | $0.55 | $1.25 | DeepSeek R2 −56% |
Output price /百万 tokens | $2.20 | $10.00 | DeepSeek R2 −78% |
Context window max input length | 128K | 1M | Gemini 2.5 Pro +7.8× |
AA Quality AA Intelligence Index (0–100) | 86 | 87 | Gemini 2.5 Pro +1pt |
Arena Elo LMArena human-pref Elo (800–2000) | — | — | Tied |
Throughput tokens per second | 110 | 140 | Gemini 2.5 Pro +27% |
P50 latency first token | 1.2s | 0.8s | Gemini 2.5 Pro −33% |
Vision multimodal | — | ||
Function calling tool use | Tied | ||
Reasoning mode chain-of-thought | Tied |
| Scenario | DeepSeek R2 | Gemini 2.5 Pro |
|---|---|---|
customer support 1000K req · 600/180 tok | $726 | $2.5K |
chat with docs 300K req · 4000/300 tok | $858 | $2.4K |
code generation 500K req · 2000/500 tok | $1.1K | $3.8K |
voice assistant 600K req · 800/200 tok | $528 | $1.8K |
Largest context window in the top tier (1M tokens). Tradeoff: pricier than DeepSeek and slower TTFT than GPT-5 — pick when whole-codebase or whole-PDF tasks matter.
Set Gemini 2.5 Pro as primary, DeepSeek R2 as fallback. One key, one bill, automatic failover when Gemini 2.5 Pro errors.
$ curl https://api.aipricly.com/v1/chat/completions \
-H "Authorization: Bearer $AIPC_KEY" \
-d '{
"routing": {
"primary": "google/gemini-2-5-pro",
"fallback": ["deepseek/deepseek-r2"]
},
"messages": [{"role": "user", "content": "..."}]
}'