Choose the right AI tool before you pay.
· 中文Start with your task, then compare the few tools that actually fit your workflow, budget, and risk level.
Artificial Analysis Intelligence Index
Which frontier model is actually smartest? Artificial Analysis aggregates 9 independent evaluations — including GPQA Diamond, Humanity’s Last Exam, and Terminal-Bench — into a single intelligence score. Higher is better.
57.653.4
52.7
48.147.5
46.4
45.4
44.844.3
43.641.8
40.9
39.537.333.7
29.2
25.0Top 20 models by Artificial Analysis Intelligence Index. Parentheses show reasoning effort; bars are scaled to the leading score from a baseline of 20.1, not zero; labels show actual scores. Data snapshot: · Index version: v4.3.2. Methodology on Artificial Analysis · See our full model rankings
Thirteen leaderboards from Artificial Analysis
The nine evaluations behind the Intelligence Index, plus per-task cost, API pricing, and output speed — the tradeoffs that decide real-world value. Higher is better except the two cost charts.
GDPval-AA v2
Quantitative analysis on real-world spreadsheets and documents.
67.3%61.7%59.8%58.7%58.4%57.3%57.0%55.3%55.0%τ³-Banking
Agentic tool use on realistic banking workflows.
50.7%50.5%50.3%48.0%47.8%47.2%47.2%46.0%44.9%41.4%SciCode
Scientific computing and research code problems.
66.9%63.1%59.5%59.0%58.8%57.6%57.4%56.6%Humanity's Last Exam
Extremely hard academic questions across dozens of domains.
61.4%59.1%54.7%48.7%47.9%47.8%46.9%43.1%GPQA Diamond
Graduate-level science questions (diamond subset).
96.1%95.3%94.9%93.7%93.5%93.5%92.9%92.8%91.7%91.2%CritPt
Critical-point reasoning under ambiguous instructions.
31.7%31.7%30.9%30.6%29.7%24.9%23.4%19.4%AA-LCR
Long-context reasoning across extended documents.
88.7%85.3%84.7%84.0%83.7%83.3%83.3%83.0%Cost per Task
Blended USD cost per Intelligence Index task, split by token type.
$0.055$0.068$0.124$0.253$0.265
$0.437$0.508
$0.549$0.607Output Speed (tok/s)
Median output tokens per second from the model API.
345288232222163149145132121104All charts: top 10 models per chart, measured independently by Artificial Analysis. Bar lengths are truncated to highlight differences; labels show actual scores. Data snapshot: . Methodology on Artificial Analysis
API pricing: cache hit, input, and output
Per-model list prices in USD per million tokens, sorted from most to least expensive by the sum of all three. Three separate bars per model; all bars share one scale, so prices are directly comparable across models.
List prices from model APIs, measured and published by Artificial Analysis. Cache-hit price is missing for models that do not offer prompt caching. Data snapshot: .
Choose by job, not by hype
Select a task to get a practical recommendation path: what to compare first, which tools are worth opening, and what risk or cost check to do before paying.
Start with Claude or ChatGPT, then check writing-specific tools only if you need workflow features.
For most writing jobs, the first decision is not “which writing app?” but “which assistant gives the best draft, edit, and reasoning loop for your budget?”
Pick the interface first: editor, extension, or terminal agent.
Cursor is strongest when the editor is the workflow. Claude Code fits terminal-heavy repo work. Plan limits and team controls matter more than raw model names.
Use citation-aware tools for discovery, then source-grounded tools for synthesis.
Perplexity is useful for sourced exploration; NotebookLM is better when you already have documents and need grounded synthesis.
Choose by output quality, control, and production constraints.
Midjourney is a strong creative benchmark, Flux is strong for model/API workflows, and Stable Diffusion-style stacks matter when control and ownership are important.
Start with the workflow: ideation, short clips, editing, or production assets.
Video tools are expensive and fast-changing, so use them for concrete creative jobs instead of subscribing before you know the output format you need.
Estimate cost before you prototype at scale.
Check input/output token pricing, free tiers, latency expectations, and fallback models before committing to one API provider.
Before buying seats, check data policy, admin controls, audit needs, and actual usage.
The right AI tool for a team is often the one with acceptable risk controls, predictable billing, and workflows people will actually use.
What are you trying to do?
These guide pages are the main entry points. Each one narrows the market to a practical shortlist.
Best AI writing tools
Choose tools for drafting, editing, long-form writing, and content workflows.
CodingBest AI coding tools
Compare assistants, code editors, terminal agents, and repo-scale help.
ImageBest AI image generators
Pick image tools by quality, control, pricing, and production use case.
VideoBest AI video generators
Understand when Runway, Sora-style tools, and video assistants make sense.
ProductivityBest AI productivity tools
Find tools for notes, meetings, tasks, docs, and day-to-day knowledge work.
ResearchBest AI tools for researchers
Compare literature search, citation-aware answers, reading, and summarization workflows.
Core comparisons with clear tradeoffs
Use these pages when you already know the category but need help picking the winner.
ChatGPT vs Claude vs Gemini
Best for choosing a daily AI assistant across writing, research, coding, and ecosystem fit.
CodingCursor vs Codex vs Claude Code
Editor, extension, and terminal-agent workflows compared by practical developer use.
ImageMidjourney vs GPT Image vs Stable Diffusion
Compare creative quality, control, model access, and production fit.
APILLM API pricing
Per-token cost comparison for building with OpenAI, Anthropic, Google, and other model providers.
PlansAI coding plan pricing
Subscription and plan comparison for AI coding tools before committing a team budget.
Ten tools we currently treat as core
These profiles support the guide and comparison pages. The site intentionally avoids pushing every long-tail tool equally.
ChatGPT
General-purpose AI assistant and reference point for most comparisons.
AssistantClaude
Strong for writing, analysis, and long-context reasoning workflows.
AssistantGemini
Google ecosystem assistant with multimodal and API use cases.
ResearchPerplexity
Useful for cited answers, web research, and quick source-backed exploration.
CodingCursor
AI-native code editor for developers who want the assistant inside the IDE.
CodingClaude Code
Terminal-oriented coding agent for repo-scale editing and planning.
ImageMidjourney
Creative image benchmark for designers, marketers, and visual exploration.
ImageFlux
Modern image model family for quality-focused generation and APIs.
VideoRunway
AI video workflow tool for creative production and short-form experiments.
ResearchNotebookLM
Research and note workflow for source-grounded reading and synthesis.
Before you choose, check cost and safety
Good AI tool decisions are not just feature decisions. Cost, limits, privacy, and workflow fit matter.
AI API pricing comparison
A guide-level view of model costs and where API pricing changes your decision.
SecurityAI tool security checklist
Questions to ask before putting business, customer, or code data into an AI product.
Lean stackBuild a smaller AI tool stack
Start with one core assistant, then add specialist tools only when a real workflow justifies them.
Track AI projects gaining momentum
Keep the picker focused on decisions, but keep a clear path to GitHub AI momentum when you want to spot fast-rising open-source projects.