AI model price-per-token comparison
What do the major AI models actually cost — and which is right for the job? Compare price per million tokens, context window, and ideal use. Sort and filter to find your fit.
| Model | Input $/1M ▾ | Output $/1M ▾ | Relative output cost | Best for |
|---|---|---|---|---|
| GPT-5.5 OpenAI · 256K ctx |
$5.00 | $15.00 | Complex reasoning & enterprise automation | |
| GPT-5 Mini OpenAI · 128K ctx |
$0.10 | $0.30 | High-volume, fast routine tasks | |
| OpenAI o-series (reasoning) OpenAI · 128K ctx |
$10.00 | $30.00 | Deep math & logical reasoning | |
| Claude Opus 4.6 Anthropic · 500K ctx |
$10.00 | $30.00 | Highest-accuracy strategic analysis | |
| Claude Sonnet 4.6 Anthropic · 500K ctx |
$2.00 | $10.00 | Balanced workhorse for coding | |
| Claude Haiku 4.6 Anthropic · 200K ctx |
$0.15 | $0.75 | Lightning-fast support bots | |
| Gemini 3 Pro Google · 2M ctx |
$2.50 | $7.50 | Massive document & video analysis | |
| Gemini 3 Flash Google · 2M ctx |
$0.20 | $0.80 | High-speed multimodal processing | |
| Llama 4 (400B) open Meta · 256K ctx |
$0.50 | $1.50 | Self-hosted, no vendor lock-in | |
| Llama 4 (70B) open Meta · 128K ctx |
$0.10 | $0.30 | Efficient edge & local deployments | |
| DeepSeek V4 open DeepSeek · 256K ctx |
$0.05 | $0.15 | Ultra-low-cost coding assistance | |
| Mistral Large 3 Mistral · 256K ctx |
$1.50 | $4.50 | Multilingual, EU data compliance |
⚠️ Prices are approximate and change frequently (last reviewed May 2026). Always confirm current pricing with each provider. "Input" = what you send the model; "output" = what it generates (usually pricier). Open-weight models can be self-hosted — shifting cost from per-token fees to your own hardware, which is exactly where our private/local AI approach comes in.
The cheapest model is rarely the right one
Token price is only part of the story — accuracy, context window, latency, and privacy all factor in. The art is matching the right model to each task so you're not overpaying for simple work or under-powering the hard stuff. That model-routing strategy is a core part of our AI consulting.
Get help choosing your stackWant this mapped to your use cases?
We'll help you pick the right model for each workflow — and show where an open or private model saves money and protects data.
Book a free Quick-Wins callConfused by the AI model maze?
We translate the alphabet soup of AI models into a clear, cost-effective stack for your business. Book a free AI Quick-Wins call.