AI Leaderboard & AGI Tracker
The strongest AI models right now, by category, plus what is coming next and a live countdown to AGI. An indicative snapshot from public leaderboards — rankings change fast and sources disagree, so always check the originals.
- Best models by category
- Upcoming models
- Live AGI countdown
AGI countdown
ARC-AGI-3 benchmark · auto-updated weeklyThe best general AI model scores about 62% on ARC-AGI-3, against the 85% human-level mark — so it has not reached human level on this test. On the current (naive) trend it might get there around February 2027. (Bespoke benchmark-specific "adapters" do hit ~99%, but at roughly $18,000 per task vs about $17 for a human, so they are excluded as they do not reflect general ability.) Frontier scores from the ARC Prize leaderboard — passing one benchmark is not general intelligence.
Best AI by category
reviewed 18 October 2026Best overall AI
General reasoning and chat (LMArena text arena).
- Gemini 4 Argon Google
- GPT-6.1 Sol OpenAI
- Claude Opus 5.5 Anthropic
- GPT-6 Astra OpenAI
Best open-source AI
Top openly-licensed models.
- Kimi K3 Moonshot AI
- DeepSeek V4 DeepSeek
- GLM-5.3 Zhipu AI
- Qwen3 Alibaba
- MiMo v2.6 Xiaomi
Best AI for coding
Code generation (LMArena WebDev / Code arena).
- Claude Opus 5.5 Anthropic
- Gemini 4 Argon Google
- GPT-6.1 Sol OpenAI
- Claude Sonnet 5.5 Anthropic
Best open-source for coding
Open-weight coding models.
- Kimi K3 Moonshot AI
- DeepSeek V4 Flash DeepSeek
- GLM-5.3 Zhipu AI
- Qwen3 Coder Alibaba
Best AI image generation
- Nano Banana Pro Google
- Seedream ByteDance
- Flux 2 Black Forest Labs
- GPT Image OpenAI
- Ideogram Ideogram
Best AI video generation
Veo 3.1 is the most consistent #1 (native audio, 4K).
- Veo 3.1 Google
- Seedance 2.0 ByteDance
- Gemini Omni Google
- Runway Runway
Best AI for speech
Text-to-speech and voice.
- ElevenLabs ElevenLabs
- Fish Speech 1.5 Fish Audio
- CosyVoice 2 Alibaba
- IndexTTS-2 Bilibili
Best AI for music
- Suno 5.5 Suno
- Lyria Google
- Stable Audio Stability AI
Best AI for science
No single public ranking; the strongest general reasoning models lead.
- Gemini 4 Argon Google
- GPT-6.1 Sol OpenAI
- Claude Opus 5.5 Anthropic
Best AI for medical
General leaders; specialist medical systems build on these. Not medical advice.
- GPT-6.1 Sol OpenAI
- Gemini 4 Argon Google
- Claude Opus 5.5 Anthropic
Best AI for legal
General leaders; specialist legal tools (e.g. Harvey, CoCounsel) build on these.
- Claude Opus 5.5 Anthropic
- GPT-6.1 Sol OpenAI
- Gemini 4 Argon Google
Best multi-function / closest to AGI
Strongest across text, image, video, audio and tools.
- Gemini 4 Argon Google
- GPT-6.1 Sol OpenAI
- Claude Opus 5.5 Anthropic
Upcoming models
rumoured / expected — not confirmed| Maker | Expected model | When | Notes |
|---|---|---|---|
| OpenAI | GPT-6.5 / GPT-7 | late 2026 to 2027 | Successor to the GPT-6 line (Astra / Sol). |
| Gemini 4.5 / 5 | 2027 | Follows Gemini 4 Argon. | |
| Anthropic | Claude Opus 6 | 2027 | Next flagship after Opus 5.5. |
| Meta | Llama 5 | 2026 to 2027 | Next open-weight flagship. |
| xAI | Grok 5 | 2026 | Next Grok iteration. |
| DeepSeek | DeepSeek V5 | 2027 | Open-weight successor to V4. |
How to read this
AI moves extremely fast and the public leaderboards often disagree on the exact order, so treat this as an indicative snapshot, not a definitive ranking. For categories like science, medical and legal there is no single public leaderboard, so the strongest general models are shown. Nothing here is professional, medical or legal advice.
The AGI countdown
ARC-AGI-3 is a benchmark designed to be easy for humans (100%) and hard for AI. The chart tracks the best public score over time and draws a simple straight-line trend to the 85% human-level mark. It is a naive projection that updates every day — not a prediction of when AGI will actually arrive.
Sources
Rankings reviewed 18 October 2026. Related: the Global Situation Monitor.