Compare 12 leading AI models side by side — price per 1M tokens, latency, context window and MMLU quality. Select the models you care about, sort by any column, export the table to Markdown or CSV, or let the recommender pick the best model for your task and budget.
Tick the models you want to compare — the table below updates in real time.
Click any column header to sort — cheapest first, fastest first, largest context or highest MMLU. Input/output prices are USD per 1M tokens; TTFT is the p50 time-to-first-token.
| Model | Input $/1M | Output $/1M | Context | p50 TTFT | MMLU | Best For |
|---|
Methodology: list prices published by each provider (Aug 2026), p50 time-to-first-token from production samples, MMLU 5-shot benchmark. Context windows in tokens (K = 1,024; 2M = 2,097,152). See Model Price Index and Model Latency Comparison for full details.
Not sure which model to start with? Tell us your task type and budget — we'll rank the 12 models and explain why.
Quick guidance for the most common choices.
GPT-5-mini ($0.15/$0.60) and DeepSeek Chat ($0.27/$1.10) deliver near-frontier quality for a fraction of flagship pricing — ideal default picks for production workloads.
Claude Sonnet 4 and GPT-5 lead on code quality with sub-450 ms latency. On a tight budget, Qwen 3.5 at $0.30/$1.20 is the standout open-weights choice.
Gemini 2.5 Pro is the only model with a 2M-token context — you can feed it an entire codebase or book corpus in one call. Claude Opus 4 (200K) and Kimi K2.6 (256K) are the strong alternatives.
DeepSeek R1 ($0.55/$2.19) gives o1-class reasoning at a fraction of the price — accept the ~900 ms first-token latency for hard math and analysis problems.
DrAI gives you one API key for all 15+ models with usage-based billing — no per-provider accounts, no setup fees. See full pricing or start free in 30 seconds.