Looking for an AI model comparison and prices for 2026? Here they all are in one table — Claude, ChatGPT (GPT), Gemini, plus the new open-weight models from China (GLM, DeepSeek, Kimi, Qwen) — with real prices, context (“memory”) size, and how much each is actually used. No marketing, just numbers.
Table: AI model prices in 2026
Sorted cheapest to most expensive (by input price):
| Model | Maker | Input /1M | Output /1M | Context | Notes |
|---|---|---|---|---|---|
| Qwen3.7 Flash | Alibaba | $0.03 | $0.13 | 1M | Dirt-cheap, fast |
| DeepSeek V4 Flash | DeepSeek | $0.09 | $0.18 | 1M | Most-used — the price killer |
| MiMo V2.5 | Xiaomi | $0.11 | $0.22 | 1M | No.1 by usage worldwide |
| MiniMax M3 | MiniMax | $0.24 | $0.96 | 1M | Cheap, high volume |
| Gemini 3.5 Flash-Lite | $0.30 | $2.50 | 1M | Google's budget option | |
| DeepSeek V4 Pro | DeepSeek | $0.44 | $0.87 | 1M | Strong & very cheap |
| GLM 5.2 | Z.ai (Zhipu) | $0.76 | $2.42 | 1M | Best open-weight model |
| Claude Haiku 4.5 | Anthropic | $1.00 | $5.00 | 200K | Fast Claude |
| GPT-5.6 Luna | OpenAI | $1.00 | $6.00 | 1M | Budget ChatGPT |
| Gemini 3.6 Flash | $1.50 | $7.50 | 1M | Balanced price/speed | |
| Claude Sonnet 5 | Anthropic | $2.00 | $10.00 | 1M | The everyday strong Claude |
| Grok 4.5 | xAI | $2.00 | $6.00 | 500K | X/Grok's model |
| GPT-5.6 Terra | OpenAI | $2.50 | $15.00 | 1M | Mid-tier GPT-5.6 |
| Kimi K3 | Moonshot | $3.00 | $15.00 | 1M | Top open model for code |
| Claude Opus 5 | Anthropic | $5.00 | $25.00 | 1M | Flagship — best at coding/agents |
| GPT-5.6 Sol | OpenAI | $5.00 | $30.00 | 1M | OpenAI's flagship |
Note: some models (e.g. GPT-5.6 Luna/Terra) often ship with promo discounts, so the price you see may be lower. “Context” is how much text the model holds at once — 1M tokens ≈ a whole book.
Which models are actually used the most?
The most honest “ranking” is not benchmarks — it is how much people actually use each one. By usage volume (weekly tokens on OpenRouter), cheap open-weight models hold the top:
- MiMo V2.5 (Xiaomi) — the most-used
- DeepSeek V4 Flash — the price killer
- GLM 5.2 (Z.ai) — the best all-round open model
- Kimi K3 (Moonshot) — strong at code
- Claude Opus 5 / Sonnet 5 — the top “Western” closed models
The pattern is clear: Chinese open-weight models win on volume(they cost almost nothing and are “good enough”), while for frontier quality in hard coding and agents, Claude remains the reference point.
Which AI model is “the best”?
There is no single winner — there is the right one for each job:
- Best quality (coding, agents): Claude Opus 5.
- Balance of price/quality: Claude Sonnet 5 or Gemini 3.6 Flash.
- Cheapest at high volume: DeepSeek V4 Flash, Qwen or MiMo.
- Best open model (privacy/self-hosting): GLM 5.2 or Kimi K3.
- Simple everyday use: ChatGPT, Claude or Gemini — whatever fits you.
Why did prices drop so much?
Two words: open-weight models. When Chinese labs released free models nearly at the level of the frontier, they forced everyone to cut prices. More in our guide on what we build with AI.
What it means for your business
You don't need to know every model. You just need to know AI got cheap and capable — and that a proper implementation costs far less than you think. See also our services.
Want us to build it for you?
Tell us what you need — we'll get back to you within 24 hours with a concrete proposal, cost and timeline.
Let's work together