Compare AI models side by side

ChatGPT, Claude, Gemini, Grok, DeepSeek and the rest, compared on what matters: how smart they are, how well they code, how much they can read at once, how fast they answer, and what they cost. Then the best model for each task, and the head-to-head comparisons people search for most.

Updated

AI model comparison table

Ranked by overall intelligence. Benchmark scores are from Artificial Analysis (May 2026); prices are per million tokens on each provider's API. Consumer subscriptions are priced differently.

ModelMakerIntelligenceCodingContextSpeedPrice in / out
GPT-5OpenAI78.476.1400K tokens142 tok/s$5.00 / $20.00
Claude Opus 4.7Anthropic76.982.31M tokens88 tok/s$15.00 / $75.00
Gemini 3 ProGoogle75.673.42M tokens165 tok/s$2.50 / $10.00
Grok 4xAI73.970.2256K tokens198 tok/s$3.00 / $15.00
DeepSeek V4 ProDeepSeek72.874.5256K tokens120 tok/s$0.27 / $1.10
Qwen 3 MaxAlibaba72.1711M tokens130 tok/s$0.50 / $2.00
Llama 4 405BMeta71.569256K tokens95 tok/s$0.90 / $2.70
Kimi K2Moonshot AI70.468.92M tokens110 tok/s$0.60 / $2.50
Mistral Large 3Mistral69.867.4256K tokens115 tok/s$2.00 / $6.00
Claude Haiku 4.5Anthropic65.866.2500K tokens205 tok/s$0.80 / $4.00
GPT-5 MiniOpenAI64.561200K tokens220 tok/s$0.25 / $1.00
Gemini 3 FlashGoogle63.260.51M tokens260 tok/s$0.15 / $0.60

How to compare AI models yourself

  1. Start from the task. A model that tops coding benchmarks can be an average writer. Use the per-task rankings above to pick two or three candidates.
  2. Use your own prompts. Paste a real piece of your work, not a riddle. That is the gap benchmarks cannot close.
  3. Send the same prompt to each. Same wording, same files, same follow-up, so the only thing that changes is the model.
  4. Weigh price and speed last. If two answers are close, the cheaper or faster model wins.

Admix does steps 3 and 4 in one go: one prompt goes to every model you pick, and the answers appear side by side.

FAQ

How do I compare AI models?

Run the same prompt through each model and judge the answers on the task you care about, then weigh price and speed. Benchmarks narrow the field, but your own prompts decide it. Admix sends one prompt to several models at once and shows the answers side by side, which makes this a single step.

Which AI model is the best right now?

No single model wins everything. Claude leads on coding and long documents, GPT models lead on reasoning and math, Gemini and Kimi handle the longest context, and DeepSeek is the cheapest at near-frontier quality. Pick by task: the "best model for each task" list on this page links to a full ranking for each one.

Which AI model is the cheapest?

Among capable models, DeepSeek and the small models (GPT-5 mini, Gemini Flash, Claude Haiku) cost the least per million tokens. The table on this page lists input and output prices for each model so you can compare them directly.

Can I compare ChatGPT, Claude and Gemini side by side?

Yes. Our head-to-head pages compare them in depth, and in Admix you can send one prompt to ChatGPT, Claude, Gemini and 300+ other models and read the answers next to each other. Admix is $8.99 a month after a 7-day trial for $1.99.

Compare them on your own prompts

One prompt, every model, side by side. A 7-day trial for $1.99, then $8.99/mo.

Try Admix for $1.99