AI Model Rankings
The AI model leaderboard, editorial.
Independent, editorial rankings of the best AI models — chat, coding, image, video, agents, audio and search. Updated weekly, translated into your language.
- models ranked
- 38
- categories
- 9
- languages
- 5
Browse by category
chat
Chat
The most contested category on the board. Chat models decide everything from everyday assistance to the hardest reasoning problems, and the gap between the best and the rest keeps shifting every week. We weigh benchmark results, real-world tool use and the human experience of long conversations to rank the models that define how most people meet AI.
→
coding
Coding
Where AI stopped being a toy and became a teammate. Coding models and agents now plan, edit and verify whole repositories, and the tools around them determine real-world velocity as much as the model behind them. We rank the complete stack — models, IDEs and autonomous agents — by how much trustworthy code they ship.
→
image
Image
Text-to-image has become the proving ground for creativity and prompt craft. These models turn a sentence into a finished image — photoreal, editorial or deliberately weird — and the best ones are now indistinguishable from a studio shoot. We rank them on aesthetics, prompt adherence and how useful the output is in real work.
→
image-editing
Image Editing
Generating an image is only half the job — the other half is fixing, filling and finishing it. Image-editing models remove backgrounds, extend scenes, retouch photos and rework shots in a single prompt. We rank the editors and their underlying models by output quality, control and how little back-and-forth they need.
→
video
Video
Text-to-video moved from a party trick to a production tool in a single year. The best models now hold physics, keep characters consistent across scenes and deliver near-cinematic quality straight from a prompt. We rank them on motion realism, control and how close the output comes to a finished shot.
→
video-editing
Video Editing
Shooting and cutting are no longer the bottleneck — editing is. Video-editing models extend clips, remove objects, translate dialogue and restructure footage in minutes instead of hours. We rank the editors by how much of the edit they can do alone and how clean the seams stay.
→
agents
Agents
The fastest-growing segment in AI, and the one that feels most like the future. Agent frameworks plan multi-step work, call tools, write code and complete tasks end to end instead of answering a single question. We rank the platforms and autonomous agents by reliability, real-world autonomy and how much you can trust them with a task.
→
audio
Audio
Voice is the interface people trust most, and text-to-speech models have crossed into broadcast quality. Today’s TTS reads with emotion, keeps a voice consistent across hours of audio and clones a voice from seconds of reference. We rank the speech models by naturalness, control and latency in real products.
→
search
Search
Search is being rebuilt from the ground up. AI answer engines now read the live web, synthesize sources and cite what they found, turning a list of links into an answer you can actually use. We rank them on answer quality, source reliability and how often they get the facts right.
→
Curated leaderboards
Benchmarks
View all →Per-benchmark AI leaderboards — GPQA, MMLU-Pro, SWE-Bench, HumanEval, LiveBench, Humanity's Last Exam, MATH and LMArena Elo. Every benchmark is normalized to one 0–100 scale and ranked from the public results published by each benchmark's maintainers.
Hugging Face
GPQA
GPQA is a multiple-choice benchmark built by Google DeepMind with 448 graduate-level science questions across physics, chemistry and biology. Its name means the questions are hard enough that search engines cannot answer them, which keeps the benchmark resistant to contamination. Scores reflect how often a model answers these expert questions correctly.
#1 GPT-5 · 99 →
Hugging Face
MMLU-Pro
MMLU-Pro is the harder successor to MMLU, the classic knowledge benchmark. It expands the original test to over 12,000 questions across 14 academic disciplines and adds ten-option answers, making it far more difficult to guess. It measures the breadth of a model’s world knowledge and how reliably it reasons across topics from law and medicine to physics and economics.
#1 GPT-5 · 98 →
BenchLM
HumanEval
HumanEval is OpenAI’s benchmark for code generation: 164 hand-written programming problems that test whether a model can produce a correct Python function from a docstring. Scores are reported as pass@1, the share of problems solved on the first attempt. It is one of the oldest and most widely cited coding benchmarks, even if real-world programming quickly outgrew it.
#1 GPT-5 · 100 →
BenchLM
SWE-Bench
SWE-Bench is the agentic coding benchmark: it gives a model a real GitHub issue from an open-source repository and asks it to produce a patch that passes the project’s tests. With more than 2,000 tasks, it measures whether an AI can navigate unfamiliar codebases, understand bug reports and land fixes that actually merge. It is the closest public proxy for a hired software engineer.
#1 GPT-5 · 93 →
LiveBench
LiveBench
LiveBench is a contamination-resistant benchmark released continuously by Abacus AI, with new question sets published on a fixed schedule so models cannot train on leaked answers. It scores 23 objective tasks across seven categories — reasoning, coding, mathematics, data analysis, language and more — where every answer is verifiable by code, not by an AI judge.
#1 GPT-5 · 99 →
BenchLM
Humanity's Last Exam
Humanity's Last Exam is the hardest knowledge benchmark ever assembled: 2,500 questions written by more than 1,000 experts across 100 subjects, from mathematics and physics to law and linguistics. Designed by the Center for AI Safety and Scale AI as a frontier test, most top models still score below 20 percent. It is the best available measure of the ceiling of AI knowledge.
#1 GPT-5 · 89 →
Hugging Face
MATH
The MATH benchmark is a collection of 5,000 competition-level mathematics problems in 7 subjects, from algebra and counting to number theory and precalculus. It tests not just calculation but step-by-step reasoning, since problems are graded on the final answer alone. It remains the standard signal for how well a model actually thinks — rather than how well it memorized.
#1 GPT-5 · 95 →
LMArena
LMArena Elo
LMArena (formerly LMSYS Chatbot Arena) is the largest human-preference leaderboard in AI. Blind side-by-side battles between models are voted on by millions of real users, and the results feed a Bradley-Terry Elo rating — the same system chess players know. It captures what people actually prefer to talk to, which no automated benchmark can fully replicate.
#1 GPT-5 · 100 →
Trending this week
#31
OpenAI AgentKit
OpenAI
97
▲ 22
OpenAI’s framework for production agent workflows — function calling, guardrails and observability built on the GPT-5 family.
#35
ElevenLabs v4
ElevenLabs
98
▲ 21
The most natural text-to-speech we test — emotion, pacing and voice cloning good enough for audiobooks, ads and agents.
#34
Manus
Manus
92
▲ 20
A general-purpose autonomous agent — given a goal, Manus plans, browses, builds and delivers the finished result.
#14
Claude Code
Anthropic
95
▲ 18
Agentic coding in the terminal — Claude plans, edits and verifies changes autonomously.