Frontier AI models compared
The most capable models available through an API, used for hard reasoning, long-horizon agentic work, and anything where a wrong answer is expensive. Prices inside this tier vary by more than 25x.
List prices checked
Token prices
| Model | Provider | Input | Output | Cached | Blended | Context |
|---|---|---|---|---|---|---|
| DeepSeek-V4-Pro | DeepSeek | $0.435 | $0.870 | $0.0036 | $0.544 | 1M |
| GPT-5 | OpenAI | $1.25 | $10.00 | $0.125 | $3.44 | 400K |
| Gemini 2.5 Pro | $1.25 | $10.00 | $0.310 | $3.44 | 1M | |
| OpenAI o3 | OpenAI | $2.00 | $8.00 | $0.500 | $3.50 | 200K |
| Mistral Large | Mistral AI | $2.00 | $6.00 | — | $3.00 | 131K |
| Cohere Command A | Cohere | $2.50 | $10.00 | — | $4.38 | 256K |
| Claude Sonnet 5 | Anthropic | $3.00 | $15.00 | $0.300 | $6.00 | 200K |
| Grok 4 | xAI | $3.00 | $15.00 | $0.750 | $6.00 | 256K |
| Claude Opus 5 | Anthropic | $5.00 | $25.00 | $0.500 | $10.00 | 200K |
| Claude Fable 5 | Anthropic | $10.00 | $50.00 | $1.00 | $20.00 | 200K |
Ranked for: In-app chat assistant
A support or help assistant handling a few thousand conversations a month, with short prompts and short answers.
| # | Model | Input cost | Output cost | Monthly |
|---|---|---|---|---|
| 1 | DeepSeek-V4-Pro | $8.70 | $4.35 | $13 |
| 2 | Mistral Large | $40 | $30 | $70 |
| 3 | GPT-5 | $25 | $50 | $75 |
| 4 | Gemini 2.5 Pro | $25 | $50 | $75 |
| 5 | OpenAI o3 | $40 | $40 | $80 |
| 6 | Cohere Command A | $50 | $50 | $100 |
| 7 | Claude Sonnet 5 | $60 | $75 | $135 |
| 8 | Grok 4 | $60 | $75 | $135 |
| 9 | Claude Opus 5 | $100 | $125 | $225 |
| 10 | Claude Fable 5 | $200 | $250 | $450 |
Ranked for: RAG over documents
Retrieval-augmented answering where each request stuffs several retrieved passages into the prompt, so input dominates.
| # | Model | Input cost | Output cost | Monthly |
|---|---|---|---|---|
| 1 | DeepSeek-V4-Pro | $131 | $13 | $144 |
| 2 | GPT-5 | $375 | $150 | $525 |
| 3 | Gemini 2.5 Pro | $375 | $150 | $525 |
| 4 | Mistral Large | $600 | $90 | $690 |
| 5 | OpenAI o3 | $600 | $120 | $720 |
| 6 | Cohere Command A | $750 | $150 | $900 |
| 7 | Claude Sonnet 5 | $900 | $225 | $1,125 |
| 8 | Grok 4 | $900 | $225 | $1,125 |
| 9 | Claude Opus 5 | $1,500 | $375 | $1,875 |
| 10 | Claude Fable 5 | $3,000 | $750 | $3,750 |
Ranked for: Coding agent
An agentic loop that reads files and writes patches, generating heavy output and re-sending large context on every turn.
| # | Model | Input cost | Output cost | Monthly |
|---|---|---|---|---|
| 1 | DeepSeek-V4-Pro | $174 | $87 | $261 |
| 2 | Mistral Large | $800 | $600 | $1,400 |
| 3 | GPT-5 | $500 | $1,000 | $1,500 |
| 4 | Gemini 2.5 Pro | $500 | $1,000 | $1,500 |
| 5 | OpenAI o3 | $800 | $800 | $1,600 |
| 6 | Cohere Command A | $1,000 | $1,000 | $2,000 |
| 7 | Claude Sonnet 5 | $1,200 | $1,500 | $2,700 |
| 8 | Grok 4 | $1,200 | $1,500 | $2,700 |
| 9 | Claude Opus 5 | $2,000 | $2,500 | $4,500 |
| 10 | Claude Fable 5 | $4,000 | $5,000 | $9,000 |