agi.jetzt
DEEN

Data last changed 15 Oct 2026, 11:38 UTC · 4 providers · 10 frontier models

How close are we
to AGI?

Which models currently count among the frontier models, and what has changed. No percentage — and where we do not know, we say so.

See AGI progressAll models

Four providers, the top model of each

  1. OpenAIGPT-6 Astrareleased 3 Sept 2026
  2. AnthropicClaude Fable 5.1released 1 Sept 2026
  3. GoogleGemini 4 Argonreleased 13 Oct 2026 · restricted access
  4. SpaceXAIGrok 4.7released 21 Sept 2026

As each provider itself positions them. All 10 frontier models at a glance ↓

What changed?

New models from these providers first, then the latest changes to models and products. All updates →

  1. GoogleModelsGemini 4 Argon releasedSource: ai.google.dev
  2. AnthropicModelsClaude Haiku 5.5 releasedSource: platform.claude.com
  3. GoogleDeploymentGemini 4 Argon is rolling out in Gemini — as reported then; current state not verifiedSource: gemini.google
  4. GoogleDeploymentGemini 4 Argon is now available in Gemini API — as reported then; current state not verifiedSource: blog.google
  5. GoogleDeploymentGemini 4 Argon is rolling out in Gemini — as reported then; current state not verifiedSource: blog.google

The frontier models

Four providers and their best models — as each provider itself positions them. Not a ranking. Checked against the providers’ pages on 4 Oct 2026.

  1. OpenAI

    Highest capabilityGPT-6 Astrareleased 3 Sept 2026our flagship model for complex reasoning and coding

    Source: developers.openai.com · checked 4 Oct

  2. Anthropic

    Highest capabilityClaude Fable 5.1released 1 Sept 2026Anthropic's most capable model open to all customers

    Source: platform.claude.com · checked 10 Oct

  3. Google

    Highest capabilityGemini 4 Argonreleased 13 Oct 2026 · restricted accessour new frontier modelAccess, per the provider: rolling out to a set of trusted cyber defenders through our Fairwind Program

    Source: blog.google · checked 10 Oct

  4. SpaceXAI

    Highest capabilityGrok 4.7released 21 Sept 2026It is the most capable model we've built

    Source: docs.x.ai · checked 4 Oct

New models · the last twelve months

When the providers released new models. That shows activity — not how far it is to AGI.

  1. OpenAIGPT-5.1 · 13 Nov 2025GPT-5.2 · 11 Dec 2025GPT-5.4 · 5 Mar 2026GPT-5.4 Pro · 5 Mar 2026GPT-5.4 mini · 17 Mar 2026GPT-5.4 nano · 17 Mar 2026GPT-5.5 · 23 Apr 2026GPT-5.5 Pro · 24 Apr 2026GPT-5.6 Luna · 9 Jul 2026GPT-5.6 Sol · 9 Jul 2026GPT-5.6 Terra · 9 Jul 2026GPT-6 Sol · 22 Sept 2026GPT-6 Astra · 3 Sept 2026GPT-6 Luna · 22 Sept 2026GPT-6.1 Sol · 29 Sept 2026GPT-6.1 Sol29 Sept
  2. AnthropicClaude Opus 4.5 · 24 Nov 2025Claude Opus 4.6 · 5 Feb 2026Claude Sonnet 4.6 · 17 Feb 2026Claude Opus 4.7 · 16 Apr 2026Claude Opus 4.8 · 28 May 2026Claude Fable 5 · 9 Jun 2026Claude Mythos 5 · 9 Jun 2026Claude Sonnet 5 · 30 Jun 2026Claude Opus 5 · 24 Jul 2026Claude Mythos 5.1 · 1 Sept 2026Claude Haiku 5.5 · 7 Oct 2026Claude Fable 5.1 · 1 Sept 2026Claude Opus 5.5 · 22 Sept 2026Claude Sonnet 5.5 · 28 Sept 2026Claude Haiku 5.57 Oct
  3. GoogleGemini 3 Pro · 18 Nov 2025Gemini 3 Flash · 17 Dec 2025Gemini 3.1 Pro · 19 Feb 2026Gemini 3.1 Flash-Lite · 7 May 2026Gemini 3.5 Flash · 19 May 2026Gemini 3.6 Flash · 21 Jul 2026Gemini 3.7 Flash · 13 Aug 2026Gemini 3.5 Flash-Lite · 21 Jul 2026Gemini 3.8 Flash · 2 Sept 2026Gemini 4 Argon · 13 Oct 2026Gemini 4 Argon13 Oct
  4. SpaceXAIGrok 4.5 · 16 Jul 2026Grok 4.6 · 12 Aug 2026Grok 4.7 · 21 Sept 2026Grok 4.721 Sept

current frontier model other language modelAll models and dates →

Only models with a documented date are entered — a release or the start of a preview. The lanes are not complete and not a comparison of the providers.

Model landscapeOutside the generalist race

AI is more than generalists

Alongside the large general-purpose models, specialized systems are emerging for images, video, speech and decisions. They are not among the frontier models and do not feed into AGI Progress – but can be important building blocks of capable AI systems.

  • Image & Design

    Generating, editing and laying out images.

    Examples
    • GPT-Image-2.5OpenAI
    • Nano Banana 2.1Google
    • FLUX 3 ImageBlack Forest Labs
    Image & Design

  • Video

    Generating moving images, often with sound.

    Examples
    • Gemini Omni FlashGoogle
    • Seedance 2.5ByteDance
    • Kling 4.0Kuaishou
    Video

  • Speech & Audio

    Understanding speech, speaking, generating music.

    Examples
    • GPT-Live 1OpenAI
    • Gemini 3.8 LiveGoogle
    • Eleven v4ElevenLabs
    Speech & Audio

  • System One & Decision Models

    Returning a choice, a score or a probability instead of text – for software, not for readers.

    1. State
    2. Decision
    3. Probability
    4. Action
    Examples
    • JevTypeSafe AI
    • Microsoft-Decision-1Microsoft
    • H2O-Lightning-4BH2O.ai
    Understand decision models →

Explore the model landscape →

AGI progress

AGI is not a single measurement. We follow ten capabilities — and show for each whether there are published results for the frontier models.

This shows what has been measured — not how good the models are, and not how far it is to AGI. There is no overall number.

See the ten capabilities

Published results on record

for all frontier models
for half of them or fewer
for none of them

AI at work

The products in which the frontier models are verified as usable today. Open Work AI →

  • OpenAI

    • ChatGPTFrontier model verified: GPT-6 Astra1 model rolling out
    • ChatGPT WorkFrontier models verified: GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna
    • CodexFrontier models verified: GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna

  • Anthropic

    • ClaudeFrontier models verified: Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5
    • Claude CodeFrontier models verified: Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5
    • Claude EnterpriseFrontier models verified: Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5
    • Claude TeamFrontier models verified: Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5

  • Google

    • GeminiFrontier models verified: Gemini 3.8 Flash, Gemini 3.5 Flash-Lite

  • GitHub

    • GitHub Copilot8 frontier models verified, from OpenAI, Anthropic, Google, SpaceXAI GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna, Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5, Gemini 3.8 Flash, Grok 4.7

  • Langdock

    • LangdockFrontier models verified: GPT-6 Luna, Claude Fable 5.1

  • Microsoft

    • Microsoft 365 Copilotmodels verified for the Cowork area onlyFrontier models verified: GPT-6 Astra, Claude Fable 5.1

No frontier model verified for: Gemini for Google Workspace, Grok. That does not mean “not available”. 13 more recorded product–model statements are not verified and not counted. “Not verified” means on record without a source check — it does not mean unavailable.

Week 41Weekly · 9 OctKI-Briefing KW 41/2026Read the issue →