GPT-4 vs Claude vs Gemini Pricing in 2026 — Real Cost Comparison

OpenAI, Anthropic, and Google charge wildly different prices for similar capability. Here is the real comparison for April 2026 — the headline rates, what they cost at common workloads, and where each provider actually wins.

Flagship Model Pricing — Head to Head

Model	Provider	Input ($/M)	Output ($/M)	Cost ratio*
GPT-4o	OpenAI	$2.50	$10.00	1.0x (baseline)
GPT-4.1	OpenAI	$2.00	$8.00	0.8x
Claude Sonnet 4	Anthropic	$3.00	$15.00	1.4x
Claude Opus 4	Anthropic	$15.00	$75.00	7.5x
Gemini 2.5 Pro	Google	$1.25	$10.00	0.85x

* Cost ratio assumes a balanced 1:1 input/output workload. GPT-4.1 is now the cheapest flagship from OpenAI. Gemini 2.5 Pro has a steep input discount but matches GPT-4o on output. Claude Sonnet 4 is the most expensive of the three "standard" flagships, and Opus 4 sits in its own pricing tier.

Cheap Tier Pricing — Head to Head

Model	Provider	Input ($/M)	Output ($/M)
GPT-4o mini	OpenAI	$0.15	$0.60
GPT-4.1 mini	OpenAI	$0.40	$1.60
GPT-4.1 nano	OpenAI	$0.10	$0.40
Claude Haiku 3.5	Anthropic	$0.80	$4.00
Gemini 2.5 Flash	Google	$0.15	$0.60
Gemini 2.0 Flash	Google	$0.10	$0.40

OpenAI and Google match each other almost exactly in the cheap tier. Anthropic's cheapest model (Haiku 3.5) is 5-8x more expensive than the OpenAI/Google equivalents — Anthropic positions Haiku as a "smarter cheap" rather than the absolute cheapest.

What These Prices Actually Cost You

Take a real workload: a customer support chatbot with 1,000 input tokens (system prompt + conversation history + user message) and 300 output tokens (assistant response), running 500 requests per day.

Model	Monthly cost	Annual cost
Gemini 2.0 Flash	$3.30	$39.60
GPT-4o mini	$4.95	$59.40
Gemini 2.5 Flash	$4.95	$59.40
Claude Haiku 3.5	$30.00	$360.00
Gemini 2.5 Pro	$63.75	$765.00
GPT-4o	$82.50	$990.00
Claude Sonnet 4	$112.50	$1,350.00
Claude Opus 4	$562.50	$6,750.00

Gemini Flash and GPT-4o mini are essentially free. Claude Haiku is cheap but 6x more than the OpenAI/Google equivalents. The flagships range from $63 to $562 per month. For a side project, this matters a lot. For a funded startup, the difference is negligible compared to engineering salaries.

Where Each Provider Actually Wins

OpenAI wins on:

• Variety — 8+ models across price tiers, easy to swap up or down
• Developer ecosystem — most tutorials, libraries, and integrations
• Function calling reliability — most mature
• Cheap tier — GPT-4o mini and 4.1 nano are both excellent

Anthropic wins on:

• Long context handling — Claude tends to use long context better than competitors
• Code quality — Sonnet and Opus often beat GPT-4o on complex coding tasks
• Long-form writing — less hedge language, fewer formulaic patterns
• Prompt caching discounts — up to 90% input reduction

Google wins on:

• Cheapest flagship — Gemini 2.5 Pro at $1.25/M input is unmatched at the top tier
• Multimodal — best video understanding, strong image handling
• Largest context window — 2M tokens on Gemini 2.5 Pro
• Free tier — Google AI Studio has the most generous free quota

How to Pick Without Picking Wrong

The honest answer: build your prompts so you can swap models easily. Use a thin abstraction layer (LiteLLM, Vercel AI SDK, or your own wrapper). Start on the cheapest model that meets quality. Upgrade individual prompts to the more expensive model only when quality forces you to.

Use the cost calculator to plug in your specific request shape and see exactly what your monthly bill looks like across every option. Then let quality, not price, drive the final pick.