AI Model Pricing Compared (November 2026): 31 Paid Models
Input, output and cached prices for 31 paid AI chat models in November 2026, with the cheapest and priciest, pricing quirks and a cost-estimate guide.
By ShilpBot Team · Published · 6 min read
Paying per token is the norm for AI models, but the prices vary by almost 300 times from the cheapest to the most expensive model. This guide compares 31 paid chat models from the OpenRouter catalogue as of 2026-11-16, explains how to estimate your real cost, and points out the pricing quirks that catch people out. Looking for free options? See our free AI chat models roundup.
AI model prices compared (sorted by output price)
| Model | Input $/1M | Output $/1M | Cached input $/1M | Context | Inputs |
|---|---|---|---|---|---|
| inclusionAI Ling 3.1 Flash | 0.045 | 0.14 | 0.009 | 1M | text |
| Ricursive Ripple 65B Preview | 0.10 | 0.30 | 0.025 | 1M | text |
| Nex AGI Nex-N2.6 | 0.12 | 0.40 | 0.024 | 262K | text, image |
| Meta Muse Ember | 0.10 | 0.50 | 0.004 | 1M | text, image, video, file, audio |
| Anthropic Claude Haiku 5.5 | 0.10 | 0.50 | 0.010 | 1M | text, image, file |
| OpenAI GPT-6 Luna | 0.10 | 0.50 | 0.010 | 1M | file, image, text |
| MiniMax M3.1 Flash | 0.15 | 0.60 | 0.015 | 1M | text, image, video |
| DeepSeek V4.1 Flash 1017 | 0.15 | 0.60 | 0.003 | 1M | text, image |
| KT Mi:dm K 3.0 Pro | 0.20 | 0.80 | 0.040 | 262K | text |
| Reflection Beam 1.0 | 0.30 | 1.00 | 0.040 | 1M | text |
| Motif Technologies Motif 3 | 0.30 | 1.20 | 0.060 | 262K | text |
| Tencent Hy4 | 0.42 | 1.25 | 0.021 | 1M | text, image |
| Z.ai GLM-5.3-FlashX-Legal | 0.37 | 1.25 | 0.075 | 1M | text, image, video |
| Cohere Command A+ | 0.30 | 1.50 | 0.15 | 192K | text, image |
| Mistral Large 4 | 0.70 | 2.10 | 0.070 | 1M | text, image |
| StepFun Step 5 Preview | 1.00 | 2.70 | 0.10 | 1M | text, image, video |
| Baidu ERNIE-Light-DT-1.0 | 1.13 | 3.94 | 0.11 | 1M | text |
| Meta Muse Spark Open | 1.25 | 4.25 | 0.15 | 1M | text, image, video, file, audio |
| Qwen4 Max | 1.50 | 4.50 | 0.075 | 1M | text, image, video |
| OpenAI GPT-6.1 Terra | 1.00 | 5.00 | 0.050 | 1M | file, image, text |
| AionLabs Aion 3.5 | 3.00 | 6.00 | 0.75 | 262K | text |
| Google Gemini 4 Hydrogen | 1.50 | 7.50 | 0.075 | 1M | audio, file, image, text, video |
| MoonshotAI Kimi K3.1 Pro | 0.24 | 8.40 | 0.20 | 1M | text, image, video |
| Z.ai GLM-5.3-Prime | 2.80 | 8.80 | 0.56 | 1M | text |
| Google Gemini 4 Carbon | 2.00 | 10.00 | 0.10 | 1M | audio, file, image, text, video |
| Google Gemini 4 Argon | 2.00 | 10.00 | 0.10 | 1M | audio, file, image, text, video |
| OpenAI GPT-6.1 Sol | 2.00 | 10.00 | 0.10 | 1M | file, image, text |
| Anthropic Claude Sonnet 5.5 | 2.00 | 10.00 | 0.10 | 1M | text, image, file |
| Qwen3.8 Max Prime | 4.00 | 12.00 | 0.50 | 1M | text, image, video |
| Meta Muse Wonder | 4.00 | 20.00 | 0.20 | 1M | text, image, video, file, audio |
| Anthropic Claude Fable 5.2 | 8.00 | 40.00 | 0.20 | 1M | text, image, file |
All prices are US dollars per million tokens. Each model has a detailed page with specs, limits, test prompts and FAQs.
The cheapest models
- inclusionAI Ling 3.1 Flash: 0.045 input, 0.14 output
- Ricursive Ripple 65B Preview: 0.10 input, 0.30 output
- Nex AGI Nex-N2.6: 0.12 input, 0.40 output
- Meta Muse Ember: 0.10 input, 0.50 output
- Anthropic Claude Haiku 5.5: 0.10 input, 0.50 output
The cheapest models suit bulk jobs such as tagging, summarising and classification, where you run thousands of items and small quality differences matter less than total cost.
The most expensive models
- Anthropic Claude Fable 5.2: 8.00 input, 40.00 output
- Meta Muse Wonder: 4.00 input, 20.00 output
- Qwen3.8 Max Prime: 4.00 input, 12.00 output
- Anthropic Claude Sonnet 5.5: 2.00 input, 10.00 output
Premium models make sense for hard problems where a better answer saves real time, such as complex debugging or high-stakes review. They are wasteful for routine tasks.
Pricing quirks to watch for
Output tokens cost more than input tokens
In every listing the output price is higher than the input price. For MoonshotAI Kimi K3.1 Pro the gap is extreme: 0.24 input against 8.40 output. Estimate your ratio of input to output before choosing.
Pro and priority variants
Several listings come in a second form. OpenAI GPT-6.1 Terra, OpenAI GPT-6.1 Sol and OpenAI GPT-6 Luna each have a Pro listing that is the same model with a deeper reasoning mode. The per-token price is the same, but pro mode spends far more tokens, so the real cost of a request is much higher. Speed variants such as Z.ai GLM-5.3-Prime and Qwen3.8 Max Prime charge more for faster output.
Cached input can be far cheaper
Most listings include a cheaper price for cached input, meaning a prompt prefix the provider has seen before. Meta Muse Ember lists cached input at 0.004 per million tokens, against 0.10 uncached. If your agent sends the same long instructions on every step, caching can dominate your savings.
A newer model can undercut an older one
Qwen4 Max costs 1.50 input and 4.50 output, which is well below Qwen3.8 Max Prime at 4.00 and 12.00. Always check whether a newer sibling gives you what you need for less.
How to estimate your real cost
- Measure a typical request: count input tokens (prompt plus context) and output tokens.
- Multiply by the model's price per million, divided by one million.
- Multiply by your expected requests per day.
- Add a margin for retries and reasoning tokens, which can double output on reasoning models.
- Re-check after a week of real use, because actual token counts usually differ from estimates.
Rules of thumb
- Start with a cheap model and move up only where it fails.
- Route simple requests to a small model and hard ones to a larger one.
- Cap output length and agent steps to avoid runaway bills.
- Check the live price before you budget, since providers change prices.
How we compiled this list
We read the public OpenRouter model catalogue on 2026-11-16 and recorded the listed input, output and cached-input prices, context window, output limit and input types for each paid chat model. We have not run benchmarks, so we do not rank models by quality. If a price has changed, tell us through the contact page.
Frequently asked questions
Which paid AI model is the cheapest?
By listed output price, inclusionAI Ling 3.1 Flash is the cheapest on our list at 0.045 input and 0.14 output per million tokens. Cheapest does not mean best, so test it on your task.
What is cached input pricing?
Cached input is a lower price for prompt text the provider has already processed. If your requests repeat the same long prefix, you pay the cheaper rate for that part.
Why do Pro versions cost more than the listed price?
Pro mode uses far more reasoning tokens per request. The per-token price is the same, but you consume many more tokens, so each request costs more.
How often do AI model prices change?
Providers change prices and add models often. We checked on 16 November 2026, so confirm the live price before you budget.
