AI Model Pricing: Every Model, Per-Token & Per-Seat
One place to see what every major AI model actually costs — per-million-token API pricing and per-seat plans, sortable and source-linked.
AI pricing comes in two shapes: usage-based API pricing, charged per million tokens of input and output, and flat per-seat plans for people using the chat apps. This page tracks both for every model we cover, with a verified date and a link to the vendor's own pricing page on every row — so you can trust the number and check it yourself.
Every Model: The Price Table
| Model↕ | Vendor↕ | Input / 1M tokens▲ | Output / 1M tokens↕ |
|---|---|---|---|
| Nova MicroAWS Bedrock on-demand | Amazon | $0.035 | $0.14 |
| Nova LiteAWS Bedrock on-demand | Amazon | $0.06 | $0.24 |
| DeepSeek V4 FlashCache-hit input $0.003 / 1M | DeepSeek | $0.14 | $0.28 |
| Mistral Small 4 | Mistral | $0.15 | $0.6 |
| Llama 4 ScoutHosted on AWS Bedrock | Meta | $0.17 | $0.66 |
| Llama 4 MaverickHosted on AWS Bedrock | Meta | $0.24 | $0.97 |
| Gemini 3.1 Flash-LiteText/image/video rate | $0.25 | $1.50 | |
| Gemini 2.5 FlashText/image/video rate | $0.3 | $2.50 | |
| CodestralCode model | Mistral | $0.3 | $0.9 |
| DeepSeek V4 ProCache-hit input $0.004 / 1M | DeepSeek | $0.435 | $0.87 |
| Mistral Large 3Flagship | Mistral | $0.5 | $1.50 |
| Command R | Cohere | $0.5 | $1.50 |
| Nova ProAWS Bedrock on-demand | Amazon | $0.8 | $3.20 |
| GPT-5.6 LunaLowest-cost tier; limited preview | OpenAI | $1 | $6 |
| Claude Haiku 4.5Fastest, cheapest Claude | Anthropic | $1 | $5 |
| Grok BuildCoding model; 256k context | xAI | $1 | $2 |
| Sonar+ search fee $5–12 / 1k requests | Perplexity | $1 | $1 |
| Gemini 2.5 Pro≤200k context; higher above | $1.25 | $10 | |
| Grok 4.3Flagship; 1M context | xAI | $1.25 | $2.50 |
| Gemini 3.5 FlashCurrent-gen Flash | $1.50 | $9 | |
| Mistral Medium 3.5 | Mistral | $1.50 | $7.50 |
| Claude Sonnet 5Intro pricing → $3 / $15 on Sep 1, 2026 | Anthropic | $2 | $10 |
| Gemini 3.1 Pro≤200k context; higher above. Preview | $2 | $12 | |
| Sonar Reasoning Pro+ per-request search fee | Perplexity | $2 | $8 |
| Sonar Deep Research+ search & citation fees | Perplexity | $2 | $8 |
| Magistral MediumReasoning | Mistral | $2 | $5 |
| GPT-5.6 TerraBalanced tier; limited preview | OpenAI | $2.50 | $15 |
| Nova PremierAWS Bedrock on-demand | Amazon | $2.50 | $12.50 |
| Command R+Command A flagship is contact-sales | Cohere | $2.50 | $10 |
| Sonar Pro+ search fee $6–14 / 1k requests | Perplexity | $3 | $15 |
| GPT-5.6 SolFlagship tier; limited preview | OpenAI | $5 | $30 |
| Claude Opus 4.8Value coder; ~half of Fable 5 | Anthropic | $5 | $25 |
| Claude Fable 5Most capable public Claude | Anthropic | $10 | $50 |
| Claude Mythos 5Shares Fable 5 pricing | Anthropic | $10 | $50 |
Per-million-token API pricing (standard tier). Some models add prompt-cache rates or per-request search fees — see the note under each. Prices last verified Jul 5, 2026 against vendor pages; always confirm current pricing on the vendor site before you buy.
Subscription & seat plans
How Token Pricing Works
Models bill per token — roughly ¾ of a word — and split the price into input (the prompt you send) and output (the text the model writes back). Output is almost always several times more expensive than input, so long, generated answers drive your bill more than long prompts do.
Cheapest vs Most Capable
The lowest per-token price is not always the lowest total cost. A cheaper model that needs more retries, longer prompts, or human correction can cost more per finished task than a pricier model that gets it right the first time. Use the table to shortlist on price, then route your hardest work to the most capable model and everyday volume to a cheaper tier.
- Lowest cost per token: Amazon Nova Micro ($0.035 / $0.14) and Nova Lite ($0.06 / $0.24).
- Cheap and capable for high volume: DeepSeek V4 Flash ($0.14 / $0.28), Mistral Small 4 ($0.15 / $0.60), Gemini 2.5 Flash ($0.30 / $2.50).
- Best value from a top lab you can deploy today: Claude Opus 4.8 ($5 / $25) and Claude Haiku 4.5 ($1 / $5).
- Most capable / premium: Claude Fable 5 ($10 / $50), GPT-5.6 Sol ($5 / $30, preview), Gemini 3.1 Pro ($2 / $12).
Estimate Your Monthly Cost
To size a workload: take your average input and output tokens per request, price each side against the table, then multiply by monthly requests. For subscription seats, multiply the per-seat price by headcount. If you are weighing automation against seats, our AI Workflow ROI Calculator turns those numbers into a payback estimate.
Plans vs API Pricing: Which Applies to You
If your team uses AI through a chat app, you pay per seat and the plan is where admin controls and data terms live. If you are building an automation or product, you pay per token through the API. Most businesses use both. Not sure which tier fits? The AI Plan & Model Picker recommends a plan and model from five quick questions.
How We Keep This Current
Every price on this page is stored in one structured source of truth and carries the date it was last verified against the vendor's official pricing page. We do not quote prices from memory. AI pricing still moves quickly, so treat this as a dated snapshot: confirm the current number on the vendor site before you commit budget, especially for newer or preview-access models, where pricing can still change.
Frequently Asked Questions
- Models are priced per million tokens, split into input (what you send) and output (what the model writes back). As of the verified date on this page, that spans a very wide range: from about $0.035 / $0.14 per million on the cheapest tier (Amazon Nova Micro) up to $10 / $50 on the most capable public models (Claude Fable 5). Output tokens almost always cost several times more than input tokens.
- Amazon Nova Micro is the cheapest on the table at $0.035 per million input and $0.14 per million output tokens, with Nova Lite ($0.06 / $0.24) and DeepSeek V4 Flash ($0.14 / $0.28) close behind. Among the big-lab flagships, GPT-5.6 Luna ($1 / $6) and Claude Haiku 4.5 ($1 / $5) are the low-cost options; Claude Opus 4.8 ($5 / $25) is the value pick when you need a top-tier model.
- GPT-5.6 Sol is cheaper per token ($5 input / $30 output) than Claude Fable 5 ($10 / $50). But GPT-5.6 is in a limited preview, while Fable 5 is generally available worldwide — so for most teams the practical comparison is between a model you can deploy now and one you may deploy later. See our full GPT-5.6 vs Claude Fable 5 comparison for the availability and compliance detail.
- A token is roughly ¾ of a word, so one million tokens is about 750,000 words. If a task sends 2,000 tokens of prompt and gets back 500 tokens on Claude Fable 5, it costs 2,000/1,000,000 × $10 + 500/1,000,000 × $50 = $0.045 — about 4.5 cents. Multiply by your monthly request volume to estimate spend, and remember output is the expensive side.
- GPT-5.6 (Sol, Terra, Luna) is in a limited preview open only to a set of vetted organizations, with general availability planned — so its preview pricing can still change. Most other models on this page, including Claude Fable 5 and Opus 4.8, are generally available and deployable today.
- Every row carries a verified date; the table above was last verified Jul 5, 2026 against vendor pricing pages. AI prices change often, so treat this as a dated snapshot and confirm the current number on the vendor site before you commit budget.
- API (per-token) pricing is usage-based — you pay for exactly the tokens you send and receive, which suits automations and products. A subscription plan (ChatGPT Plus, Claude Team, etc.) is a flat per-seat monthly fee for people using the chat app, and is where admin controls and data terms live. Most businesses use both: seats for staff, API for automation.
Not Sure Which Model Fits Your Budget?
We map the right model and plan to your actual workflows, volume, and compliance needs — and show you where to spend and where to save. Start with a free AI workflow audit.
Book a Consultation