LLMCost

Classification & Routing LLM Cost

Categorizing tickets, emails, or documents at scale. Monthly estimates at 100,000 requests/mo, avg 600 input / 50 output tokens per request.

#ModelEst. monthly costvs cheapest
1Qwen: Qwen3.7 Flash$2.45—
2DeepSeek: DeepSeek V4.1 Flash (batch)$3.751.5×
3Anthropic: Claude Haiku 5.5 (batch)$4.251.7×
4OpenAI: GPT-6 Luna Pro (batch)$4.251.7×
5OpenAI: GPT-6 Luna (batch)$4.251.7×
6DeepSeek: DeepSeek V4.1 Flash$4.822.0×
7OpenAI: GPT-5 Nano$5.002.0×
8Google: Gemini 2.5 Flash Lite$8.003.3×
9Anthropic: Claude Haiku 5.5$8.503.5×
10OpenAI: GPT-6 Luna Pro$8.503.5×
11OpenAI: GPT-6 Luna$8.503.5×
12Qwen: Qwen3.8 27B$9.153.7×
13Qwen: Qwen3.8 Omni Flash$11.354.6×
14Qwen: Qwen3.8 Flash$11.354.6×
15DeepSeek: DeepSeek V4.1 Flash 1017$12.004.9×
16Meta: Llama 4 Maverick$14.515.9×
17DeepSeek: DeepSeek V4 Flash Vision Exp$16.176.6×
18OpenAI: GPT-5.6 Luna$18.007.3×
19OpenAI: GPT-5.4 Nano$18.257.4×
20DeepSeek: DeepSeek V3.1$19.758.1×
21OpenAI: GPT-5 Mini$25.0010.2×
22Google: Gemini 3.5 Flash Lite$30.5012.4×
23Google: Gemini 2.5 Flash$30.5012.4×
24Google: Gemini 3.8 Flash (batch)$31.8813.0×
25Google: Gemini 3.7 Flash (batch)$31.8813.0×
26DeepSeek: DeepSeek V4 Flash 0731$33.0013.5×
27DeepSeek: DeepSeek V4 Pro 0813$49.5020.2×
28Google: Gemini 3.7 Flash$63.7526.0×
29Google: Gemini 3.6 Flash$63.7526.0×
30Amazon: Nova Pro 1.0$64.0026.1×
31OpenAI: GPT-5.4 Mini$67.5027.6×
32Google: Gemini 3.8 Flash$70.1328.6×
33Google: Gemini 4 Argon (batch)$85.0034.7×
34OpenAI: GPT-6.1 Sol Pro (batch)$85.0034.7×
35OpenAI: GPT-6.1 Sol (batch)$85.0034.7×
36Anthropic: Claude Sonnet 5.5 (batch)$85.0034.7×
37OpenAI: GPT-6 Sol Pro (batch)$85.0034.7×
38OpenAI: GPT-6 Sol (batch)$85.0034.7×
39OpenAI: o4 Mini$88.0035.9×
40Mistral: Mistral Large 4$102.5041.8×
41OpenAI: GPT-5$125.0051.0×
42Google: Gemini 2.5 Pro$125.0051.0×
43Google: Nano Banana 2.1$127.5052.0×
44Google: Gemini 3.5 Flash$135.0055.1×
45xAI: Grok 4.7$150.0061.2×
46Qwen: Qwen3.8 Max (0902)$150.0061.2×
47Qwen: Qwen3.8 2.4T A95B$150.0061.2×
48xAI: Grok 4.6$150.0061.2×
49OpenAI: o3$160.0065.3×
50Anthropic: Claude Sonnet 5$170.0069.4×
51Google: Gemini 4 Argon$170.0069.4×
52OpenAI: GPT-6.1 Sol Pro$170.0069.4×
53OpenAI: GPT-6.1 Sol$170.0069.4×
54Anthropic: Claude Sonnet 5.5$170.0069.4×
55OpenAI: GPT-6 Sol Pro$170.0069.4×
56OpenAI: GPT-6 Sol$170.0069.4×
57Anthropic: Claude Opus 5.5 (batch)$170.0069.4×
58OpenAI: GPT-5.6 Terra Pro$180.0073.5×
59Anthropic: Claude Opus 5 (batch)$212.5086.7×
60OpenAI: GPT-5.4$225.0091.8×
61Anthropic: Claude Sonnet 4.5$255.00104.1×
62Qwen: Qwen3.8 Max Prime$300.00122.4×
63OpenAI: GPT-5.6 Sol$340.00138.8×
64Anthropic: Claude Opus 5.5$340.00138.8×
65Anthropic: Claude Opus 5$425.00173.5×
66OpenAI: GPT-6 Astra (batch)$425.00173.5×
67OpenAI: GPT-6 Astra Pro (batch)$425.00173.5×
68Anthropic: Claude Fable 5.1 (batch)$425.00173.5×
69OpenAI: GPT-5.5$450.00183.7×
70Anthropic: Claude Fable 5$850.00346.9×
71OpenAI: GPT-6 Astra$850.00346.9×
72OpenAI: GPT-6 Astra Pro$850.00346.9×
73Anthropic: Claude Fable 5.1$850.00346.9×
74Anthropic: Claude Opus 4.1$1275.00520.4×
75OpenAI: GPT-5 Pro$1500.00612.2×
76OpenAI: GPT-5.5 Pro$2700.001102.0×

Cost-saving tips for classification & routings

  • Extreme volume × tiny output = cheapest possible model wins outright.
  • Even small per-request savings compound massively at 100K requests/mo.
  • Use logprobs or structured outputs to make routing deterministic.

Customize the numbers with the interactive calculator · All scenarios