LLM API pricing — cost per point of intelligenceLLM API 价格——每美元能买多少智能
Every tracked model with its API list price and a value metric: value = Intelligence Score ÷ blended price (blended = average of input and output price per 1M tokens). Higher value = more intelligence per dollar. Sorted by value by default; click any column to re-sort. Prices sync from the OpenRouter API daily (refreshed 2026-09-12).
所有被追踪模型的 API 标价与性价比指标:性价比 = 智能评分 ÷ 混合价(混合价 = 输入输出每百万 token 价格的均值)。性价比越高,每美元买到的智能越多。默认按性价比排序,点表头可换列。价格每日从 OpenRouter API 同步(更新于 2026-09-12)。
Best value right now: DeepSeek V4 Flash 0731 (1473.3), Qwen3.7 Flash (1020), DeepSeek V4 Flash 0423 (832.5) (score per blended $).当前性价比前三:DeepSeek V4 Flash 0731 (1473.3), Qwen3.7 Flash (1020), DeepSeek V4 Flash 0423 (832.5)(每混合美元换来的分数)。
| # | Model模型 | Score总分 | $ in/1M输入$/1M | $ out/1M输出$/1M | Blended混合价 | Value性价比 | Ctx上下文 |
|---|---|---|---|---|---|---|---|
| 1 | DeepSeek V4 Flash 0731DeepSeek | $0.040 | $0.080 | $0.060 | 1473.3 | 1.3M | |
| 2 | Qwen3.7 Flash(on/off开/关)Qwen | $0.030 | $0.130 | $0.080 | 1020 | 1M | |
| 3 | DeepSeek V4 Flash 0423(max)DeepSeek | $0.066 | $0.131 | $0.099 | 832.5 | 1M | |
| 4 | gpt-oss-120bOpenAI | $0.037 | $0.170 | $0.104 | 611.6 | 131K | |
| 5 | Hy3Tencent | $0.083 | $0.330 | $0.206 | 394.2 | 262K | |
| 6 | Phi 4Microsoft | $0.070 | $0.140 | $0.105 | 355.2 | 16K | |
| 7 | Seed-2.0-MiniByteDance Seed | $0.100 | $0.400 | $0.250 | 310.8 | 262K | |
| 8 | DeepSeek V3.2(on/off开/关)DeepSeek | $0.269 | $0.400 | $0.335 | 209 | 164K | |
| 9 | Mistral Small 4Mistral AI | $0.150 | $0.600 | $0.375 | 129.3 | 262K | |
| 10 | GPT-5.6 Luna(max)OpenAI | $0.200 | $1.20 | $0.700 | 119 | 1.1M | |
| 11 | Step 3.7 FlashStepFun | $0.200 | $1.15 | $0.675 | 116.7 | 262K | |
| 12 | Llama 4 ScoutMeta | $0.100 | $0.300 | $0.200 | 115 | 1.3M | |
| 13 | Mercury 2Inception | $0.250 | $0.750 | $0.500 | 114.4 | 128K | |
| 14 | MiMo-V2.5-ProXiaomi | $0.435 | $0.870 | $0.653 | 111.6 | 1.1M | |
| 15 | MiniMax M3MiniMax | $0.300 | $1.20 | $0.750 | 106.4 | 1M | |
| 16 | Qwen3.7 Plus(on/off开/关)Qwen | $0.320 | $1.28 | $0.800 | 102.1 | 1M | |
| 17 | MiniMax M2.7MiniMax | $0.300 | $1.20 | $0.750 | 99.2 | 205K | |
| 18 | GPT-5.6 Luna Pro(max)OpenAI | $0.200 | $1.20 | $0.700 | 98.9 | 1.1M | |
| 19 | Llama 4 MaverickMeta | $0.200 | $0.696 | $0.448 | 91.7 | 1M | |
| 20 | ERNIE 4.5 VL 424B A47B Baidu | $0.420 | $1.25 | $0.835 | 80.2 | 123K | |
| 21 | GLM 5.2(max)Z.ai | $0.600 | $2.00 | $1.30 | 66.4 | 1M | |
| 22 | Gemini 3.1 Flash Lite(high)Google | $0.250 | $1.50 | $0.875 | 63.7 | 1M | |
| 23 | Seed-2.0-LiteByteDance Seed | $0.250 | $2.00 | $1.13 | 62.3 | 262K | |
| 24 | Kimi K2 Thinking(always-on常开)Moonshot AI | $0.600 | $2.50 | $1.55 | 45.2 | 262K | |
| 25 | Grok 4.20(high)xAI | $1.25 | $2.50 | $1.88 | 42.7 | 2M | |
| 26 | Nemotron 3 UltraNVIDIA | $0.625 | $3.13 | $1.88 | 39.7 | 262K | |
| 27 | Gemini 3.5 Flash Lite(high)Google | $0.300 | $2.50 | $1.40 | 39.1 | 1M | |
| 28 | Grok 4.3(high)xAI | $1.25 | $2.50 | $1.88 | 38.6 | 1M | |
| 29 | Kimi K2.7 CodeMoonshot AI | $0.710 | $3.50 | $2.10 | 36.3 | 262K | |
| 30 | DeepSeek V4 Pro 0423(max)DeepSeek | $1.60 | $3.20 | $2.40 | 36 | 1M | |
| 31 | Gemini 3.6 Flash(high)Google | $0.750 | $3.75 | $2.25 | 36 | 1M | |
| 32 | GLM 5.1(on/off开/关)Z.ai | $0.966 | $3.04 | $2.00 | 35.5 | 205K | |
| 33 | Qwen3.7 Max(on/off开/关)Qwen | $1.48 | $4.42 | $2.95 | 30 | 1M | |
| 34 | Muse Spark 1.2meta | $1.25 | $4.25 | $2.75 | 29 | 1M | |
| 35 | Muse Spark 1.1meta | $1.25 | $4.25 | $2.75 | 25.2 | 1M | |
| 36 | Qwen3.8 Max (0902)Qwen | $2.00 | $6.00 | $4.00 | 22.6 | 1M | |
| 37 | Grok 4.5(high)xAI | $2.00 | $6.00 | $4.00 | 20.1 | 500K | |
| 38 | Claude Haiku 4.5(high)Anthropic | $1.00 | $5.00 | $3.00 | 17 | 200K | |
| 39 | GPT-5.6 Sol(max)OpenAI | $2.00 | $10 | $6.00 | 15.6 | 1.1M | |
| 40 | Gemini 3.5 Flash(high)Google | $1.50 | $9.00 | $5.25 | 15.3 | 1M | |
| 41 | GPT-5.6 Sol Pro(max)OpenAI | $2.00 | $10 | $6.00 | 14.2 | 1.1M | |
| 42 | Mistral Medium 3.5Mistral AI | $1.50 | $7.50 | $4.50 | 14.2 | 262K | |
| 43 | Claude Sonnet 5(high)Anthropic | $2.00 | $10 | $6.00 | 13.9 | 1M | |
| 44 | GPT-5.6 Terra(max)OpenAI | $2.00 | $12 | $7.00 | 12.6 | 1.1M | |
| 45 | Nova Pro 1.0Amazon | $0.800 | $3.20 | $2.00 | 12.6 | 300K | |
| 46 | Mistral LargeMistral AI | $2.00 | $6.00 | $4.00 | 12.2 | 128K | |
| 47 | Gemini 3.1 Pro Preview(high)Google | $2.00 | $12 | $7.00 | 11.8 | 1M | |
| 48 | Kimi K3(always-on常开)Moonshot AI | $2.65 | $13 | $7.96 | 10.7 | 1M | |
| 49 | Command ACohere | $2.50 | $10 | $6.25 | 7.9 | 256K | |
| 50 | Command R+ (08-2024)Cohere | $2.50 | $10 | $6.25 | 6.8 | 128K | |
| 51 | Claude Opus 5(max)Anthropic | $5.00 | $25 | $15 | 5.9 | 1M | |
| 52 | Sonar ProPerplexity | $3.00 | $15 | $9.00 | 5.8 | 200K | |
| 53 | Claude Opus 4.8(max)Anthropic | $5.00 | $25 | $15 | 5.6 | 1M | |
| 54 | GPT-5.5(max)OpenAI | $5.00 | $30 | $18 | 4.7 | 1.1M | |
| 55 | Nova Premier 1.0Amazon | $2.50 | $13 | $7.50 | 4.5 | 1M | |
| 56 | Claude Fable 5(max)Anthropic | $10 | $50 | $30 | 3.1 | 1M |
Models without a public score or without list prices are omitted from the value table — see the leaderboard for the full catalog.未出分或无价模型不进性价比表——完整目录见排行榜。
Monthly cost calculator月成本计算器
Estimate each model at your own usage, then set the minimum score you can accept.按你的实际用量估算每个模型的月账单,再设定能接受的最低分。
How the score works →评分方法 → · List prices via the official OpenRouter API; volume discounts, cached-token pricing and batch APIs not included.价格为 OpenRouter 官方 API 标价,未含阶梯折扣、缓存命中价与批量 API。