GPT vs Claude vs Gemini: pricing compared
OpenAI's GPT, Anthropic's Claude and Google's Gemini are the three families most teams weigh up first. All three price per token with separate input and output rates, and all three offer several tiers. Exact numbers change often, so this guide focuses on how each lineup is structured and how to read the trade-offs — then points you to a live tool for today's figures.
💡 Just want the current numbers? Open the comparison tool, pick any two models, and see input/output prices, context and a monthly cost estimate side by side.
How each family is structured
OpenAI — GPT
OpenAI typically offers the widest tier spread: small "mini"/"nano" models priced for high volume, a mid flagship that's the default for most production work, and a premium "pro" tier for the hardest reasoning. That range means GPT can be both one of the cheaper and one of the pricier options depending on which model you pick.
Anthropic — Claude
Claude's lineup is usually a clean three-step ladder: a small fast model (Haiku), a balanced workhorse (Sonnet), and a premium model (Opus). Anthropic tends to hold list prices steadier than rivals but leans on discounts — prompt caching and a batch API — to lower effective cost. Claude is often favoured for coding and long-context work.
Google — Gemini
Gemini is frequently the value play: its "Flash" and "Flash-Lite" tiers sit among the lowest paid prices in the market, it offers a notably generous free tier, and it tends to advertise very large context windows. The premium "Pro" tier competes with the other flagships.
What actually moves your bill
- Which tier, not which brand. The gap between a family's cheap and premium model is usually far bigger than the gap between brands at the same tier. Matching the tier to the task matters more than the logo.
- Output price. Output is billed at a higher rate than input across all three, and often dominates the total — so weigh output price heavily for chat and generation.
- Discounts. Caching and batch processing can cut effective cost dramatically and aren't reflected in the headline number.
- Context you actually send. A huge context window is only cheap if you don't fill it on every call.
How to choose
Pick by task and budget, then verify on the live data. Use the calculator with a workload preset to rank all three families by estimated monthly cost for your token mix, compare your two finalists head to head, and check each provider's page for its full model list. Then apply the tactics in how to cut your LLM API costs to bring the bill down further.
The short version
GPT offers the widest range, Claude a clean three-tier ladder with steady pricing plus strong discounts, and Gemini the most aggressive value and free tier. But the tier you choose and your output length will move your bill more than the brand. Compare on the live table for your specific usage rather than trusting a fixed ranking.
Model names and prices change frequently. This article describes general 2026 market structure, not fixed rates — always confirm current pricing on each provider's own page and use the live TokenSwarm calculator and comparison tool. TokenSwarm is independent and not affiliated with OpenAI, Anthropic, Google or any provider.