For this week's AI for GTM newsletter, we're looking at how AI model pricing has changed over the last 12 months. We compared OpenAI, Anthropic, and Google to see what happened to their flagship and lower-cost models, and whether AI is actually getting cheaper. |
|
AI model pricing got a lot more complicated |
|
AI model pricing used to be pretty easy to follow. Newest model = expensive. But in the last 12 months, that relationship has gotten a lot messier. Providers have launched pricier flagships, added cheaper variants, and started pricing things like cached context and longer prompts differently. So we looked at three of the biggest names in AI, OpenAI, Anthropic, and Google, to see what actually changed. Here’s what we found 👇 |
|
OpenAI went deep on multi-tier pricing strategy |
|
OpenAI has pushed its frontier model prices way up. GPT-6 Astra now costs $10 per million input tokens and $50 per million output tokens. Compared with OpenAI’s top model from a year ago, that’s an 8× increase in input pricing and a 5× increase in output pricing. But OpenAI isn't expecting everyone to pay flagship prices |
|
Claude is pushing its advanced models higher while keeping lower tiers competitive |
|
Anthropic has taken a similar approach for its front model. Its newest flagship, Fable 5.1, now costs $10/$50, doubling both input and output pricing compared to Claude Opus 4.5 launched at $5/$25 per million input/output tokens in November 2025. |
|
But prices lower down the lineup have moved in the other direction. Sonnet 4.5 and Sonnet 4.6 were priced at $3/$15 during launch but its upgraded Sonnet 5 is now $2/$10, a 33% reduction in both input and output pricing. Anthropic has also been making cached contexts much cheaper. Fable 5.1 cache reads cost $0.25 per million tokens, while regular input costs $10. That means cached reads cost just 2.5% of regular input. |
|
Gemini keeps expanding the Flash family |
|
This year, Google released Gemini 3 Flash, Gemini 3.6 Flash, Gemini 3.7 Flash, and Gemini 3.8 Flash. Gemini 3 Flash costs $0.50 per million input tokens and $3 per million output tokens. Compared with Gemini 2.5 Pro at $1.25/$10, that’s a 60% drop in input pricing and a 70% drop in output pricing. Google then moved through the Flash lineup with Gemini 3.6, 3.7, and 3.8. The newer models are currently priced at $0.75/$3.75, which is 50% more expensive on input and 25% more expensive on output than Gemini 3 Flash. Even with that increase, the newer Flash models remain cheaper than Gemini 2.5 Pro, while taking on more complex work like coding, agentic tasks, and multi-step execution. |
|
So, what trends are we noticing? |
|
Looking across the AI market over the last year, here are some of our understanding: Frontier models are getting more expensive The biggest providers are putting higher prices on their most capable models. OpenAI and Anthropic have both pushed their latest flagships into much more expensive price tiers. But at the same time, cheaper models like from Deepseek have started putting pressure on those prices. When DeepSeek launched models at a fraction of what many frontier models cost, it showed that capable AI didn't always have to come with a premium price tag. That put more pressure on other providers to give customers cheaper options. And that's where the cheaper variants come in. Instead of having one model at one price, providers are building wider model lineups. You can pay more for the most capable model, or choose a cheaper variant that can still handle coding, reasoning, agents, or everyday workloads. |
|
Want to see what happened across the rest of the AI market?
We compared seven major AI providers over the last 12 months, including DeepSeek, Grok, Mistral, and Kimi.
| |
|
| That's all, folks! I'll see you soon with more tips and ideas. If there is something we can do better for you, please let me know by replying to this email. |
|
Until next time,
Aquib
CEO, Mailmodo |
|
|
|
