Tool
AI model API cost calculator
Enter daily requests, average input and output tokens and cache hit rate to compare per-request and monthly cost across major models.
Usage inputs
Monthly cost ranking
| # | Model | Context | Per request | Per month | vs cheapest |
|---|---|---|---|---|---|
| 1 | DeepSeek V4-FlashDeepSeek | 1M | $0.0006 | $168.5 | base |
| 2 | Claude Haiku 5.5Anthropic | 1M | $0.0006 | $193.8 | ×1.2 |
| 3 | GPT-6 LunaOpenAI | 1.1M | $0.0006 | $193.8 | ×1.2 |
| 4 | Qwen3.8-FlashAlibaba | 992K | $0.0007 | $223.7 | ×1.3 |
| 5 | MiniMax-M3MiniMax | 1M | $0.0017 | $514.8 | ×3.1 |
| 6 | Gemini 3.1 Flash-LiteGoogle | 1M | $0.0018 | $544.5 | ×3.2 |
| 7 | Mistral Large 3Mistral AI | 262K | $0.0024 | $729.0 | ×4.3 |
| 8 | DeepSeek V4-ProDeepSeek | 1M | $0.0024 | $731.4 | ×4.3 |
| 9 | Gemini 3 FlashGoogle | 1M | $0.0036 | $1,089 | ×6.5 |
| 10 | GLM-5.3Zhipu Z.ai | 1M | $0.0070 | $2,111 | ×12.5 |
| 11 | Qwen3.8-MaxAlibaba | 992K | $0.0097 | $2,925 | ×17.4 |
| 12 | Grok 4.7xAI | 500K | $0.0099 | $2,970 | ×17.6 |
| 13 | Claude Sonnet 5.5Anthropic | 1M | $0.013 | $3,858 | ×22.9 |
| 14 | GPT-6.1 SolOpenAI | 922K | $0.013 | $3,858 | ×22.9 |
| 15 | GPT-6 SolOpenAI | 1.1M | $0.013 | $3,876 | ×23.0 |
| 16 | Gemini 3.1 ProGoogle | 1M | $0.015 | $4,356 | ×25.9 |
| 17 | GPT-5.2OpenAI | 272K | $0.016 | $4,651 | ×27.6 |
| 18 | Kimi K3Moonshot AI | 1M | $0.019 | $5,814 | ×34.5 |
| 19 | Claude Opus 5.5Anthropic | 1M | $0.026 | $7,716 | ×45.8 |
| 20 | Claude Fable 5.1Anthropic | 1M | $0.064 | $19,245 | ×114.2 |
| 21 | GPT-6 AstraOpenAI | 1.1M | $0.065 | $19,380 | ×115.0 |
Cache-hit tokens are billed at each model's cache-read price; models without a cache price bill all input at the standard rate. Estimates exclude cache writes, batch discounts, long-context tiers and free quotas.
Official price table
Verified 2026-10-08| Vendor | Model | Input | Cached input | Output | Context | Source |
|---|---|---|---|---|---|---|
| DeepSeek | DeepSeek V4-Flash | $0.140 | $0.0028 | $0.280 | 1M | Official* |
| Anthropic | Claude Haiku 5.5 | $0.100 | $0.010 | $0.500 | 1M | Official |
| OpenAI | GPT-6 Luna | $0.100 | $0.010 | $0.500 | 1.1M | Official |
| Alibaba | Qwen3.8-Flash | $0.150 | $0.016 | $0.470 | 992K | Official* |
| MiniMax | MiniMax-M3 | $0.300 | $0.060 | $1.20 | 1M | Official* |
| Gemini 3.1 Flash-Lite | $0.250 | $0.025 | $1.50 | 1M | Official | |
| Mistral AI | Mistral Large 3 | $0.500 | $0.050 | $1.50 | 262K | Official |
| DeepSeek | DeepSeek V4-Pro | $0.435 | $0.0036 | $1.74 | 1M | Official* |
| Gemini 3 Flash | $0.500 | $0.050 | $3.00 | 1M | Official* | |
| Zhipu Z.ai | GLM-5.3 | $1.40 | $0.260 | $4.40 | 1M | Official |
| Alibaba | Qwen3.8-Max | $2.00 | $0.250 | $6.00 | 992K | Official* |
| xAI | Grok 4.7 | $2.00 | $0.500 | $6.00 | 500K | Official |
| Anthropic | Claude Sonnet 5.5 | $2.00 | $0.100 | $10.0 | 1M | Official* |
| OpenAI | GPT-6.1 Sol | $2.00 | $0.100 | $10.0 | 922K | Official* |
| OpenAI | GPT-6 Sol | $2.00 | $0.200 | $10.0 | 1.1M | Official* |
| Gemini 3.1 Pro | $2.00 | $0.200 | $12.0 | 1M | Official* | |
| OpenAI | GPT-5.2 | $1.75 | $0.175 | $14.0 | 272K | Official |
| Moonshot AI | Kimi K3 | $3.00 | $0.300 | $15.0 | 1M | Official |
| Anthropic | Claude Opus 5.5 | $4.00 | $0.200 | $20.0 | 1M | Official* |
| Anthropic | Claude Fable 5.1 | $10.0 | $0.250 | $50.0 | 1M | Official |
| OpenAI | GPT-6 Astra | $10.0 | $1.00 | $50.0 | 1.1M | Official |
- DeepSeek V4-Flash：deepseek-chat/reasoner aliases retired 2026-07; this is the successor model
- Qwen3.8-Flash：International USD pricing
- MiniMax-M3：Standard tier with permanent 50% discount; rates double above 512K input
- DeepSeek V4-Pro：Priced at 1/4 of list after the 75%-off promotion ended 2026-05-31
- Gemini 3 Flash：Output price includes thinking tokens
- Qwen3.8-Max：International USD pricing; mainland China prices are listed in CNY separately
- Claude Sonnet 5.5：Cache reads billed at 5% of base input price
- GPT-6.1 Sol：Requests with >272K input tokens bill input/cache at 2x and output at 1.5x
- GPT-6 Sol：Long-context tier applies above 272K input tokens per request
- Gemini 3.1 Pro：Prompts >200K tokens bill $4 in / $18 out; grounding billed separately
- Claude Opus 5.5：Cache writes billed separately (~1.25x input for 5m); batch API is 50% off
How we compute
Monthly cost = daily requests × 30 × (input tokens × effective input price + output tokens × output price) ÷ 1,000,000. The effective input price blends the standard input price and the cache-read price at your cache hit rate. Output price covers the model's generated tokens, including thinking tokens where the vendor bills them as output.
All figures come from each vendor's official pricing page and are re-checked by hand; the verification date is shown above the table. Real invoices can differ: long-context tiers, batch and priority lanes, cache write and storage fees, free quotas and region-specific pricing are not modeled here.
FAQ
What counts as a request?
One API call. If your app makes N calls a day, enter N as daily requests. For streaming chat, one conversation turn usually equals one call.
How does prompt caching lower cost?
Repeated prefixes such as system prompts, tool definitions and retrieved context can be served from cache at a much lower read price. A higher hit rate shifts input spend toward the cheaper rate; models without a published cache price gain no discount.
Why is my real bill different?
This tool uses list prices only. Long-context tiers, batch lanes, cache write and storage fees, grounding tools, free quotas and regional pricing all change the final number. Check the linked official page before budgeting.
How often are prices updated?
The table is maintained by hand against official pricing pages and shows its last verification date. Vendors change prices often, so treat the numbers as a comparison baseline, not a quote.