A VS Code fork embedded with powerful frontier models that can write, refactor, and debug code across your entire repository with multi-file awareness.

Multi-file editsRepo-wide indexTab autocompletionClaude 3.5 Sonnet

Explore Cursor AI

LLMs & Writing

PopularFreemium

Claude 3.5 Sonnet

4.9

The industry-leading model for reasoning, coding and complex nuance

Anthropic flagship intelligence model known for superior coding benchmarks, deep nuanced writing, and the interactive Artifacts workspace.

Interactive Artifacts200K token contextLeading coding benchmarksVision

Explore Claude 3.5 Sonnet

LLMs & Writing

PopularFreemium

ChatGPT (GPT-4o)

4.8

The versatile multimodal intelligence engine by OpenAI

Omni-model integrating voice, vision, and text in real time. Powers advanced web browsing, Python code execution, and custom GPT assistants.

Voice modeAdvanced Data AnalysisDALL·E 3 Image GenerationCustom GPTs

Explore ChatGPT (GPT-4o)

Productivity & Agents

TrendingFreemium

Perplexity AI

4.9

Conversational answer engine with direct citations from the live web

Replaces standard search engines with synthesized answers, cited sources, real-time follow-ups, and multi-file document analysis.

Real-time web searchInline source citationsFocus modesPro search

Explore Perplexity AI

Image & Art

PopularPaid

Midjourney v6

4.9

Photorealistic image generation and artistic visual synthesis

The gold standard in text-to-image AI, renowned for exceptional texture realism, coherent lighting, fine typographic generation, and web canvas editing.

Photorealism v6Inpainting & OutpaintingWeb interfaceStyle reference

Explore Midjourney v6

Image & Art

HotFreemium

FLUX.1 by Black Forest Labs

4.8

Open-weight generative imaging model redefining text rendering and anatomy

Cutting-edge 12B parameter flow matching visual model rivaling Midjourney with state-of-the-art hand anatomy, prompt adherence, and open weights.

Open-weight variantsFlawless text in imagesPhotorealistic handsFast inference

Explore FLUX.1 by Black Forest Labs

Benchmark Data

Frontier Models — Side-by-Side

Compare context windows, coding benchmarks, reasoning scores, and API pricing across providers.

Frontier Model	Context Window	SWE-bench (Code)	HumanEval	Pricing	Ideal Use Case
Claude 3.7 Sonnet Top Score Anthropic	200,000 tokens	70.3% (Verified)	93.7%	$3 / $15 per 1M tokens	Hybrid reasoning with thinking mode, complex multi-file code editing, and precise refactoring.
GPT-4o OpenAI	128,000 tokens	38.8%	90.2%	$2.50 / $10 per 1M tokens	Real-time voice, vision analysis, structured outputs, and general conversational intelligence.
DeepSeek R1 DeepSeek	128,000 tokens	49.2%	92.8%	$0.55 / $2.19 per 1M tokens	Deep chain-of-thought reasoning, mathematical proofs, and competitive programming.
Gemini 2.0 Flash Google	1,000,000 tokens	24.1%	84.1%	$0.10 / $0.40 per 1M tokens	Ultra-fast inference, massive document processing, and cost-efficient high-throughput workloads.

How We Score, Measure, and Price Every Model

Published so you can audit any number on this site — no black-box composite scores.

How is the intelligence index calculated?

The intelligence index is a composite 0–100 score that blends ten independent evaluations: SWE-bench Verified for agentic coding, GPQA Diamond for graduate-level science, MMLU-Pro for broad knowledge, AIME for mathematical reasoning, HumanEval and LiveCodeBench for code generation, Terminal-Bench for shell tool use, Humanity’s Last Exam for frontier reasoning, long-context retrieval, and LMArena Elo for human preference. Each benchmark is normalised against the current frontier maximum before weighting, so a model cannot inflate its score with a single narrow strength.

How do you measure output speed?

Speed is measured as steady-state output tokens per second through each provider’s official API, sampled over twenty consecutive requests at a fixed 1,000-token prompt. We additionally publish time-to-first-token, because latency-sensitive chat interfaces care far more about the first token than about sustained throughput. Both figures are re-measured monthly since providers change serving infrastructure without notice.

How is cost per task computed?

Cost per task models a realistic agentic workload rather than a single round trip. We assume a 45,000-output-token task with a 10,000-token input context, apply each model’s published input and output prices per million tokens, and add reasoning-token overhead for models that expose a thinking mode. Batch and cached-input discounts are excluded so the figure reflects the default interactive path most teams actually pay.

Where does the benchmark data come from?

Every figure is sourced from provider release specifications, official model cards, and independently reproduced benchmark runs. We do not republish third-party leaderboard numbers without re-running the evaluation ourselves, and each model page cites its primary source and last-verified date so readers can audit any claim.

How often is the leaderboard updated?

Pricing and availability are checked weekly. Benchmark scores are re-run whenever a provider ships a point release that changes model behaviour, and at minimum once per month. The updated date shown on the leaderboard always reflects the most recent full re-evaluation, not the last cosmetic edit.

Careers & Immigration