LLMCost

Coding Agent LLM Cost

Cursor/Claude Code style dev assistants with large contexts. Monthly estimates at 5,000 requests/mo, avg 8,000 input / 2,000 output tokens per request.

#ModelEst. monthly costvs cheapest
1Qwen: Qwen3.7 Flash$2.50—
2DeepSeek: DeepSeek V4.1 Flash (batch)$3.501.4×
3Anthropic: Claude Haiku 5.5 (batch)$4.501.8×
4OpenAI: GPT-6 Luna Pro (batch)$4.501.8×
5OpenAI: GPT-6 Luna (batch)$4.501.8×
6DeepSeek: DeepSeek V4.1 Flash$5.712.3×
7OpenAI: GPT-5 Nano$6.002.4×
8Google: Gemini 2.5 Flash Lite$8.003.2×
9Anthropic: Claude Haiku 5.5$9.003.6×
10OpenAI: GPT-6 Luna Pro$9.003.6×
11OpenAI: GPT-6 Luna$9.003.6×
12Qwen: Qwen3.8 Omni Flash$10.704.3×
13Qwen: Qwen3.8 Flash$10.704.3×
14DeepSeek: DeepSeek V4.1 Flash 1017$12.004.8×
15Meta: Llama 4 Maverick$14.035.6×
16DeepSeek: DeepSeek V4 Flash Vision Exp$15.096.0×
17Qwen: Qwen3.8 27B$15.106.0×
18DeepSeek: DeepSeek V3.1$19.507.8×
19OpenAI: GPT-5.6 Luna$20.008.0×
20OpenAI: GPT-5.4 Nano$20.508.2×
21OpenAI: GPT-5 Mini$30.0012.0×
22DeepSeek: DeepSeek V4 Flash 0731$30.8012.3×
23Google: Gemini 3.8 Flash (batch)$33.7513.5×
24Google: Gemini 3.7 Flash (batch)$33.7513.5×
25Google: Gemini 3.5 Flash Lite$37.0014.8×
26Google: Gemini 2.5 Flash$37.0014.8×
27DeepSeek: DeepSeek V4 Pro 0813$46.2018.5×
28Amazon: Nova Pro 1.0$64.0025.6×
29Google: Gemini 3.7 Flash$67.5027.0×
30Google: Gemini 3.6 Flash$67.5027.0×
31Google: Gemini 3.8 Flash$74.2529.7×
32OpenAI: GPT-5.4 Mini$75.0030.0×
33OpenAI: o4 Mini$88.0035.2×
34Google: Gemini 4 Argon (batch)$90.0036.0×
35OpenAI: GPT-6.1 Sol Pro (batch)$90.0036.0×
36OpenAI: GPT-6.1 Sol (batch)$90.0036.0×
37Anthropic: Claude Sonnet 5.5 (batch)$90.0036.0×
38OpenAI: GPT-6 Sol Pro (batch)$90.0036.0×
39OpenAI: GPT-6 Sol (batch)$90.0036.0×
40Mistral: Mistral Large 4$96.2038.5×
41Google: Nano Banana 2.1$135.0054.0×
42xAI: Grok 4.7$140.0056.0×
43Qwen: Qwen3.8 Max (0902)$140.0056.0×
44Qwen: Qwen3.8 2.4T A95B$140.0056.0×
45xAI: Grok 4.6$140.0056.0×
46Google: Gemini 3.5 Flash$150.0060.0×
47OpenAI: GPT-5$150.0060.0×
48Google: Gemini 2.5 Pro$150.0060.0×
49OpenAI: o3$160.0064.0×
50Anthropic: Claude Sonnet 5$180.0072.0×
51Google: Gemini 4 Argon$180.0072.0×
52OpenAI: GPT-6.1 Sol Pro$180.0072.0×
53OpenAI: GPT-6.1 Sol$180.0072.0×
54Anthropic: Claude Sonnet 5.5$180.0072.0×
55OpenAI: GPT-6 Sol Pro$180.0072.0×
56OpenAI: GPT-6 Sol$180.0072.0×
57Anthropic: Claude Opus 5.5 (batch)$180.0072.0×
58OpenAI: GPT-5.6 Terra Pro$200.0080.0×
59Anthropic: Claude Opus 5 (batch)$225.0090.0×
60OpenAI: GPT-5.4$250.00100.0×
61Anthropic: Claude Sonnet 4.5$270.00108.0×
62Qwen: Qwen3.8 Max Prime$280.00112.0×
63OpenAI: GPT-5.6 Sol$360.00144.0×
64Anthropic: Claude Opus 5.5$360.00144.0×
65Anthropic: Claude Opus 5$450.00180.0×
66OpenAI: GPT-6 Astra (batch)$450.00180.0×
67OpenAI: GPT-6 Astra Pro (batch)$450.00180.0×
68Anthropic: Claude Fable 5.1 (batch)$450.00180.0×
69OpenAI: GPT-5.5$500.00200.0×
70Anthropic: Claude Fable 5$900.00360.0×
71OpenAI: GPT-6 Astra$900.00360.0×
72OpenAI: GPT-6 Astra Pro$900.00360.0×
73Anthropic: Claude Fable 5.1$900.00360.0×
74Anthropic: Claude Opus 4.1$1350.00540.0×
75OpenAI: GPT-5 Pro$1800.00720.0×
76OpenAI: GPT-5.5 Pro$3000.001200.0×

Cost-saving tips for coding agents

  • Input dominates cost here (8K in vs 2K out) — prompt caching is your biggest lever.
  • Use a frontier model only for planning steps; route file edits to mid-tier.
  • Trim context: send relevant files, not the whole repo.

Customize the numbers with the interactive calculator · All scenarios