LLMCost

Data Extraction LLM Cost

Structured JSON extraction from unstructured text. Monthly estimates at 15,000 requests/mo, avg 2,500 input / 300 output tokens per request.

#ModelEst. monthly costvs cheapest
1Qwen: Qwen3.7 Flash$1.71—
2DeepSeek: DeepSeek V4.1 Flash (batch)$2.551.5×
3Anthropic: Claude Haiku 5.5 (batch)$3.001.8×
4OpenAI: GPT-6 Luna Pro (batch)$3.001.8×
5OpenAI: GPT-6 Luna (batch)$3.001.8×
6DeepSeek: DeepSeek V4.1 Flash$3.522.1×
7OpenAI: GPT-5 Nano$3.672.1×
8Google: Gemini 2.5 Flash Lite$5.553.2×
9Anthropic: Claude Haiku 5.5$6.003.5×
10OpenAI: GPT-6 Luna Pro$6.003.5×
11OpenAI: GPT-6 Luna$6.003.5×
12Qwen: Qwen3.8 27B$7.584.4×
13Qwen: Qwen3.8 Omni Flash$7.744.5×
14Qwen: Qwen3.8 Flash$7.744.5×
15DeepSeek: DeepSeek V4.1 Flash 1017$8.324.9×
16Meta: Llama 4 Maverick$9.975.8×
17DeepSeek: DeepSeek V4 Flash Vision Exp$11.006.4×
18OpenAI: GPT-5.6 Luna$12.907.5×
19OpenAI: GPT-5.4 Nano$13.137.7×
20DeepSeek: DeepSeek V3.1$13.658.0×
21OpenAI: GPT-5 Mini$18.3810.7×
22DeepSeek: DeepSeek V4 Flash 0731$22.4413.1×
23Google: Gemini 3.5 Flash Lite$22.5013.2×
24Google: Gemini 2.5 Flash$22.5013.2×
25Google: Gemini 3.8 Flash (batch)$22.5013.2×
26Google: Gemini 3.7 Flash (batch)$22.5013.2×
27DeepSeek: DeepSeek V4 Pro 0813$33.6619.7×
28Amazon: Nova Pro 1.0$44.4026.0×
29Google: Gemini 3.7 Flash$45.0026.3×
30Google: Gemini 3.6 Flash$45.0026.3×
31OpenAI: GPT-5.4 Mini$48.3828.3×
32Google: Gemini 3.8 Flash$49.5028.9×
33Google: Gemini 4 Argon (batch)$60.0035.1×
34OpenAI: GPT-6.1 Sol Pro (batch)$60.0035.1×
35OpenAI: GPT-6.1 Sol (batch)$60.0035.1×
36Anthropic: Claude Sonnet 5.5 (batch)$60.0035.1×
37OpenAI: GPT-6 Sol Pro (batch)$60.0035.1×
38OpenAI: GPT-6 Sol (batch)$60.0035.1×
39OpenAI: o4 Mini$61.0535.7×
40Mistral: Mistral Large 4$69.8140.8×
41Google: Nano Banana 2.1$90.0052.6×
42OpenAI: GPT-5$91.8853.7×
43Google: Gemini 2.5 Pro$91.8853.7×
44Google: Gemini 3.5 Flash$96.7556.6×
45xAI: Grok 4.7$102.0059.6×
46Qwen: Qwen3.8 Max (0902)$102.0059.6×
47Qwen: Qwen3.8 2.4T A95B$102.0059.6×
48xAI: Grok 4.6$102.0059.6×
49OpenAI: o3$111.0064.9×
50Anthropic: Claude Sonnet 5$120.0070.2×
51Google: Gemini 4 Argon$120.0070.2×
52OpenAI: GPT-6.1 Sol Pro$120.0070.2×
53OpenAI: GPT-6.1 Sol$120.0070.2×
54Anthropic: Claude Sonnet 5.5$120.0070.2×
55OpenAI: GPT-6 Sol Pro$120.0070.2×
56OpenAI: GPT-6 Sol$120.0070.2×
57Anthropic: Claude Opus 5.5 (batch)$120.0070.2×
58OpenAI: GPT-5.6 Terra Pro$129.0075.4×
59Anthropic: Claude Opus 5 (batch)$150.0087.7×
60OpenAI: GPT-5.4$161.2594.3×
61Anthropic: Claude Sonnet 4.5$180.00105.3×
62Qwen: Qwen3.8 Max Prime$204.00119.3×
63OpenAI: GPT-5.6 Sol$240.00140.4×
64Anthropic: Claude Opus 5.5$240.00140.4×
65Anthropic: Claude Opus 5$300.00175.4×
66OpenAI: GPT-6 Astra (batch)$300.00175.4×
67OpenAI: GPT-6 Astra Pro (batch)$300.00175.4×
68Anthropic: Claude Fable 5.1 (batch)$300.00175.4×
69OpenAI: GPT-5.5$322.50188.6×
70Anthropic: Claude Fable 5$600.00350.9×
71OpenAI: GPT-6 Astra$600.00350.9×
72OpenAI: GPT-6 Astra Pro$600.00350.9×
73Anthropic: Claude Fable 5.1$600.00350.9×
74Anthropic: Claude Opus 4.1$900.00526.3×
75OpenAI: GPT-5 Pro$1102.50644.7×
76OpenAI: GPT-5.5 Pro$1935.001131.6×

Cost-saving tips for data extractions

  • Structured outputs / JSON mode eliminates parsing failures — use it.
  • Extraction is quality-sensitive per field; sample-audit before scaling down.
  • Batch APIs shine here: latency rarely matters for pipeline jobs.

Customize the numbers with the interactive calculator · All scenarios