Model Comparison
Compare AI language models side by side: context window, pricing, capabilities, and speed. Runs on your device — nothing is uploaded, nothing is stored.
| Model ↑ | Provider | Context Window | Input (1M tokens) | Output (1M tokens) | Tier | Multimodal |
|---|---|---|---|---|---|---|
| Claude Haiku 4.5 | Anthropic | 200,000 | $0.8 | $4 | Fast | Yes |
| Claude Opus 4.6 | Anthropic | 200,000 | $15 | $75 | Flagship | Yes |
| Claude Sonnet 4.6 | Anthropic | 200,000 | $3 | $15 | Balanced | Yes |
| DeepSeek V3 | DeepSeek | 64,000 | $0.27 | $1.1 | Balanced | No |
| Gemini 2.0 Flash | 1,000,000 | $0.1 | $0.4 | Fast | Yes | |
| Gemini 2.0 Pro | 2,000,000 | $1.25 | $5 | Flagship | Yes | |
| GPT-4o | OpenAI | 128,000 | $2.5 | $10 | Flagship | Yes |
| GPT-4o mini | OpenAI | 128,000 | $0.15 | $0.6 | Fast | Yes |
| Llama 3.1 405B | Meta (via Together) | 128,000 | $3.5 | $3.5 | Open Source | No |
| Llama 3.3 70B | Meta (via Groq) | 128,000 | $0.59 | $0.79 | Open Source | No |
| Mistral Large | Mistral | 128,000 | $2 | $6 | Balanced | No |
| o1 | OpenAI | 200,000 | $15 | $60 | Flagship | No |
| o3-mini | OpenAI | 200,000 | $1.1 | $4.4 | Fast | No |
Claude Haiku 4.5
Anthropic
Context Window
200,000 tokens
Multimodal
Yes
Input (1M tokens)
$0.8
Output (1M tokens)
$4
Claude Opus 4.6
Anthropic
Context Window
200,000 tokens
Multimodal
Yes
Input (1M tokens)
$15
Output (1M tokens)
$75
Claude Sonnet 4.6
Anthropic
Context Window
200,000 tokens
Multimodal
Yes
Input (1M tokens)
$3
Output (1M tokens)
$15
DeepSeek V3
DeepSeek
Context Window
64,000 tokens
Multimodal
No
Input (1M tokens)
$0.27
Output (1M tokens)
$1.1
Gemini 2.0 Flash
Context Window
1,000,000 tokens
Multimodal
Yes
Input (1M tokens)
$0.1
Output (1M tokens)
$0.4
Gemini 2.0 Pro
Context Window
2,000,000 tokens
Multimodal
Yes
Input (1M tokens)
$1.25
Output (1M tokens)
$5
GPT-4o
OpenAI
Context Window
128,000 tokens
Multimodal
Yes
Input (1M tokens)
$2.5
Output (1M tokens)
$10
GPT-4o mini
OpenAI
Context Window
128,000 tokens
Multimodal
Yes
Input (1M tokens)
$0.15
Output (1M tokens)
$0.6
Llama 3.1 405B
Meta (via Together)
Context Window
128,000 tokens
Multimodal
No
Input (1M tokens)
$3.5
Output (1M tokens)
$3.5
Llama 3.3 70B
Meta (via Groq)
Context Window
128,000 tokens
Multimodal
No
Input (1M tokens)
$0.59
Output (1M tokens)
$0.79
Mistral Large
Mistral
Context Window
128,000 tokens
Multimodal
No
Input (1M tokens)
$2
Output (1M tokens)
$6
o1
OpenAI
Context Window
200,000 tokens
Multimodal
No
Input (1M tokens)
$15
Output (1M tokens)
$60
o3-mini
OpenAI
Context Window
200,000 tokens
Multimodal
No
Input (1M tokens)
$1.1
Output (1M tokens)
$4.4
This page lines up the large language models people actually choose between in one sortable table: OpenAI's GPT-4o, GPT-4o mini, o1 and o3-mini; Anthropic's Claude Opus, Sonnet and Haiku; Google's Gemini Flash and Pro; Llama 3.3 70B and Llama 3.1 405B as served by third-party hosts; Mistral Large; and DeepSeek V3. For each one you get the context window in tokens, the price per million input tokens, the price per million output tokens, a tier, and whether the model is multimodal.
Type in the search box to find a model or provider by name, narrow the table with the provider and tier filters, and click the Model, Context, Input price or Output price headings to sort, clicking again to reverse the order. Tiers are this site's own grouping: flagship, balanced, fast, and open for the open-weight Llama models. On a phone the table turns into one card per model.
Why this runs on your device
There is nothing to protect here: the table is public pricing data and you are not entering any text. The tool simply runs in the page, so filtering and sorting are instant and need no account or API key.
- It compares price, context size, tier and a multimodal flag, not quality. There are no benchmark scores, speed or latency figures, and nothing tells you which model is best at your task.
- The figures come from a fixed table kept in the site's code. They are list prices per million tokens at the time the table was last updated and will not follow a provider's price changes on their own; check the provider's own pricing page before you commit to a budget.
- Only text-token list prices are shown. Cached-input discounts, batch pricing, long-context surcharges and the cost of images or audio are not included.
- The context window is the maximum a model accepts, not what it can generate in one reply, and the multimodal column is a simple yes or no rather than a list of supported input types.
- Llama prices are those of a specific host (Groq for 70B, Together for 405B, as named in the provider column), so they will differ from other hosts or from running the model yourself.