LLM Model Comparison
Compare features, pricing, and capabilities of the latest language models side-by-side — Claude Opus 4.8, GPT-5.5, Gemini 3 Pro, Grok 4, Llama 4, and more.
Pricing per 1M tokens (USD). Last updated June 2026.
Model | Provider | Context | Max Output | Input Price | Output Price | Vision | Functions | JSON Mode |
|---|---|---|---|---|---|---|---|---|
Claude Haiku 4.5 2025-10 | Anthropic | 200,000 | 64,000 | $1.00 per 1M tokens | $5.00 per 1M tokens | |||
Claude Opus 4.8 2026-05 | Anthropic | 1,000,000 | 128,000 | $5.00 per 1M tokens | $25.00 per 1M tokens | |||
Claude Sonnet 4.6 2025-11 | Anthropic | 1,000,000 | 64,000 | $3.00 per 1M tokens | $15.00 per 1M tokens | |||
DeepSeek V3.2 2025-12 | DeepSeek | 128,000 | 8,192 | $0.28 per 1M tokens | $0.42 per 1M tokens | |||
Gemini 3 Flash 2026-03 | 1,000,000 | 65,536 | $0.50 per 1M tokens | $3.00 per 1M tokens | ||||
Gemini 3 Flash-Lite 2026-03 | 1,000,000 | 65,536 | $0.10 per 1M tokens | $0.40 per 1M tokens | ||||
Gemini 3 Pro 2025-12 | 1,000,000 | 65,536 | $2.00 per 1M tokens | $12.00 per 1M tokens | ||||
GPT-5 mini 2025-08 | OpenAI | 400,000 | 128,000 | $0.25 per 1M tokens | $2.00 per 1M tokens | |||
GPT-5 nano 2025-08 | OpenAI | 400,000 | 128,000 | $0.05 per 1M tokens | $0.40 per 1M tokens | |||
GPT-5.1 2025-11 | OpenAI | 400,000 | 128,000 | $1.25 per 1M tokens | $10.00 per 1M tokens | |||
GPT-5.5 2026-04 | OpenAI | 400,000 | 128,000 | $5.00 per 1M tokens | $30.00 per 1M tokens | |||
Grok 4 2025-07 | xAI | 256,000 | 64,000 | $3.00 per 1M tokens | $15.00 per 1M tokens | |||
Grok 4 Fast 2025-09 | xAI | 256,000 | 64,000 | $0.20 per 1M tokens | $0.50 per 1M tokens | |||
Llama 4 Maverick 2025-04 | Meta | 1,000,000 | 16,384 | $0.35 per 1M tokens | $1.15 per 1M tokens | |||
Llama 4 Scout 2025-04 | Meta | 1,000,000 | 16,384 | $0.11 per 1M tokens | $0.34 per 1M tokens | |||
Mistral Large 3 2025-11 | Mistral | 256,000 | 16,384 | $2.00 per 1M tokens | $6.00 per 1M tokens |
How to choose the right LLM
There is no single best language model — only the best fit for a given task, budget, and latency target. A frontier model such as Claude Opus 4.8 or GPT-5.5 earns its premium on hard reasoning, long agentic runs, and nuanced writing. For classification, extraction, summarization, and everyday chat, a fast and inexpensive model like Gemini 3 Flash or Claude Haiku 4.5 often delivers near-identical results at a fraction of the cost and latency.
This comparison table lines up the specifications that actually drive that decision: input and output price, context window, maximum output length, and whether the model supports vision, function calling, and structured JSON output. Sorting by any column makes trade-offs obvious — for example, the cheapest model with a one-million-token context window, or the fastest model that still handles images.
How to use this tool
- Filter by provider, or leave it on all to compare every model side by side.
- Click a column header to sort by price, context window, or output length.
- Match capabilities (vision, function calling, JSON mode) to your use case before optimizing for price.
Frequently asked questions
What is a context window and why does it matter?+
It's the maximum number of tokens — system prompt, conversation, and any retrieved context — the model can consider at once. Bigger windows let you feed in more documents or longer histories, but cost scales with how much you actually use.
Do I always need the most capable model?+
No. Most production traffic is routine and runs well on a mid-tier or small model. A common pattern is to route simple requests to a cheap model and escalate only the hard ones to a frontier model.
What does function calling / JSON mode give me?+
They let the model return structured, machine-readable output reliably — essential for tool use, agents, and pipelines where you parse the response programmatically rather than showing raw text to a user.