🤖 Updated: July 20, 2026

AI Models in 2026: comparison table

Q: How often is this table updated?

With every site content update; prices are always checked against the three labs' official pages.

Q: Why aren't all existing models listed?

We curate the active, relevant models in each tier from the three major labs so the table stays useful at a glance.

Q: Do these prices include batch or caching discounts?

No, these are standard API prices. Batch typically offers 50% off and prompt caching can cut repeated-token costs by up to 10x.

Every current model from Anthropic, OpenAI and Google in one table: API price per million tokens, tier and context window. Verified against the official pricing pages.

Model	Lab	Tier	Input $/M	Output $/M	Context
Claude Fable 5	Anthropic	Frontier	10.00	50.00	1M
Claude Opus 4.8	Anthropic	Advanced reasoning	5.00	25.00	1M
Claude Sonnet 5	Anthropic	Balanced	2.00*	10.00*	1M
Claude Haiku 4.5	Anthropic	Fast & affordable	1.00	5.00	1M
GPT-5.5-pro	OpenAI	Frontier	30.00	180.00	—
GPT-5.6-sol	OpenAI	High-end	5.00	30.00	—
GPT-5.5	OpenAI	High-end	5.00	30.00	1M
GPT-5.6-terra	OpenAI	Balanced	2.50	15.00	—
GPT-5.6-luna	OpenAI	Fast & affordable	1.00	6.00	—
GPT-5.4-nano	OpenAI	Ultra-budget	0.20	1.25	—
Gemini 3.1 Pro (preview)	Google	High-end	2.00**	12.00**	—
Gemini 3.5 Flash	Google	Balanced	1.50	9.00	—
Gemini 2.5 Flash	Google	Fast & affordable	0.30	2.50	1M
Gemini 3.1 Flash-Lite	Google	Ultra-budget	0.25	1.50	—

* Sonnet 5 promotional pricing through August 31, 2026; $3/$15 per million afterwards. ** Tier up to 200K tokens; above that, $4/$18. Dashes mean the lab doesn't publish a context window on its pricing page. Standard API prices in USD, without batch or caching discounts, verified on July 20, 2026 against the official pages of Anthropic, OpenAI and Google.

How to read this table

The most expensive model isn't always the best one for your case. Frontier models (Fable 5, GPT-5.5-pro) shine at deep reasoning, long agentic tasks and research — but for 80% of everyday tasks, a balanced model like Sonnet 5, GPT-5.6-terra or Gemini 3.5 Flash delivers excellent results at a fraction of the price. And for classifying, extracting or answering simple queries at high volume, the budget tier (Haiku 4.5, GPT-5.4-nano, Flash-Lite) is unbeatable on cost per result.

Also note the input/output asymmetry: output costs 5–6× more than input on nearly every model. If your use case generates long responses (writing, code), the output price weighs far more on your bill than the input price.

From data to decision

With these prices in front of you, the next step is plugging in your own numbers: our cost calculator estimates your monthly spend from your real volume, the model selector tells you in 5 questions which Claude fits you, and the Claude vs ChatGPT vs Gemini comparison analyzes the differences beyond price.

// FAQ

Frequently asked questions

How often is this table updated?

With every site content update — the exact date of the last verification is shown in the badge at the top and in the note under the table. Prices are always checked against the three labs' official pages, never third-party sources.

Why aren't all existing models listed?

We curate the active, relevant models in each tier (frontier, high-end, balanced and budget) from the three major labs. Older versions that are still available but superseded (like Opus 4.5–4.7 or Sonnet 4.5) are omitted so the table stays useful at a glance.

Do these prices include batch or caching discounts?

No — these are standard API prices. Batch processing typically offers ~50% off and prompt caching can cut the cost of repeated tokens by up to 10×. Our cost calculator helps you estimate the impact of those savings.