The AI Model Pricing Index
Across the 8 frontier models we’ve tracked since May 2026, the typical input price has fallen 33%. Here is what every major model costs per million tokens today, and how those prices are moving.
Input and output API price per 1M tokens. Verified daily against provider pricing by our model tracker. Free to reuse with attribution (CC BY 4.0). Download CSV · JSON
Recent price changes
- 2026-09-23Gemini Omni Flashinput$1.5→$0.15/1M-90%
- 2026-09-23Gemini Omni Flashoutput$17.5→$0.6/1M-97%
- 2026-09-09Gemini Omni Flashinput$1.5→$0.15/1M-90%
- 2026-09-09Gemini Omni Flashoutput$17.5→$0.6/1M-97%
- 2026-09-04Cohere Parse 5output$1500→$1.5/1M-100%
- 2026-09-02Claude Fable 5output$50→$0.25/1M-99%
- 2026-07-02Claude Opus 4.7input$15→$5/1M-67%
- 2026-07-02Claude Opus 4.7output$75→$25/1M-67%
- 2026-07-02Claude Haiku 4.5input$0.8→$1/1M+25%
- 2026-07-02Claude Haiku 4.5output$4→$5/1M+25%
- 2026-07-02GPT-5input$10→$1.25/1M-87%
- 2026-07-02GPT-5output$30→$10/1M-67%
- 2026-07-02Gemini 2.5 Flashinput$0.075→$0.3/1M+300%
- 2026-07-02Mistral Largeinput$3→$2/1M-33%
- 2026-07-02Mistral Largeoutput$9→$6/1M-33%
- 2026-07-02Claude Opus 4.8input$15→$5/1M-67%
Current pricing (61 models, cheapest input first)
| Model | Input /1M | Output /1M |
|---|---|---|
| North Mini CodeCohere | $0 | $0 |
| Ox AlphaAnonymous (Stealth) | $0 | $0 |
| JevTypeSafe AI | $0.042 | Open |
| GLM-5.3-FlashZ.ai (Zhipu AI) | $0.075 | $0.25 |
| Gemini Omni FlashGoogle | $0.15 | $0.6 |
| Gemini 2.5 FlashGoogle DeepMind | $0.3 | $2.5 |
| Gemini 3.7 FlashGoogle | $0.75 | $3.75 |
| Gemini 3.8 FlashGoogle | $0.75 | $3.75 |
| Amazon Nova ProAmazon Web Services | $0.8 | $3.2 |
| Claude Haiku 4.5Anthropic | $1 | $5 |
| GPT-5OpenAI | $1.25 | $10 |
| Gemini 2.5 ProGoogle DeepMind | $1.25 | $10 |
| GLM-5.2Zhipu AI / Z.ai | $1.4 | $4.4 |
| Gemini 3.5Google DeepMind | $1.5 | $9 |
| Gemini 3.6 FlashGoogle | $1.5 | $7.5 |
| Gemini 3.5 Flash CyberGoogle | $1.5 | $7.5 |
| Mistral LargeMistral AI | $2 | $6 |
| Grok 4.5SpaceXAI (xAI) | $2 | $6 |
| Qwen 3.8 MaxAlibaba | $2 | $6 |
| Gemini Robotics ER 2Google DeepMind | $2 | $10 |
| Palmyra X6Writer | $2 | $8 |
| Grok 4.6SpaceXAI | $2 | $6 |
| Grok 4.7xAI (SpaceXAI) | $2 | $6 |
| GPT-4oOpenAI | $2.5 | $10 |
| Command R+Cohere | $2.5 | $10 |
| GPT-5.4OpenAI | $2.5 | $15 |
| Gemini 3.5 TranscribeGoogle | $2.5 | $12 |
| Claude Sonnet 4.6Anthropic | $3 | $15 |
| Claude Sonnet 5Anthropic | $3 | $15 |
| Kimi K3Moonshot AI | $3 | $15 |
| Claude Opus 5.5Anthropic | $4 | $20 |
| GPT-5.5OpenAI | $5 | $30 |
| Claude Opus 4.8Anthropic | $5 | $25 |
| Claude Opus 4.7Anthropic | $5 | $25 |
| GPT-5.6OpenAI | $5 | $30 |
| GPT-5.6 SolOpenAI | $5 | $30 |
| Claude Opus 5Anthropic | $5 | $25 |
| Claude Fable 5Anthropic | $10 | $0.25 |
| AstraOpenAI | $10 | $50 |
| GPT-6 AstraOpenAI | $10 | $50 |
| Claude Fable 5.1Anthropic | $10 | $50 |
| Claude Mythos 5.1Anthropic | $10 | $50 |
| GPT-5.6-CyberOpenAI | $12.5 | $75 |
| o1OpenAI | $15 | $60 |
| Antares-1BCisco | Open | Open |
| Cohere Parse 5Cohere | Open | Open |
| DeepSeek V3DeepSeek | Open | Open |
| DeepSeek V4DeepSeek | Open | Open |
| DeepSeek V4 FlashDeepSeek | Open | Open |
| Gemma 3Google DeepMind | Open | Open |
| Gemma 4 12BGoogle | Open | Open |
| Laguna S 2.1Poolside | Open | Open |
| Llama 4Meta | Open | Open |
| MAI-Cyber-1-FlashMicrosoft | Open | Open |
| Qwen 3Alibaba (Qwen Team) | Open | Open |
| Qwen 3.8 27BAlibaba | Open | Open |
| Qwen 3.8-Flash-NextAlibaba | Open | Open |
| Robostral NavigateMistral AI | Open | Open |
| Salesforce KoaSalesforce | Open | Open |
| SWE-2Cognition | Open | Open |
| ThomsonThomson Reuters | Open | Open |
Methodology & reuse
Prices are the public list price per 1 million tokens for each model’s standard API tier, re-verified daily by our automated model tracker. Open-weight models are marked “Open” (self-hostable; no fixed per-token list price). A new data point is recorded only when a price actually changes, so every entry in the changes feed is a real event.
This dataset is free to reuse with attribution to AIToolCrunch (CC BY 4.0). Grab the CSV or JSON.