SUN, 19 JUL 2026 · 00:04:15 UTC
ai.newsdeck
Department · The Timeline414 curated events

A short history
of modern AI.

Browse by day, month, or year. Scrub through every month of activity in the sparkline below — or filter by category to follow a single thread.

AllModelResearchProductPolicyFundingIndustryIncidentMilestone
Oct 2018Jul 2026

July

2026

3 events

  1. Jul 18

    2026

    model

    TestingCatalog: Kimi K3 Hits AI/ML API, Benchmarked Against Fable 5

    Kimi K3 is available on AI/ML API alongside Claude Fable 5, with vibe-coding benchmarks showing K3 outperforming Fable 5 at a quarter of the price.

    Source

  2. Jul 18

    2026

    model

    TestingCatalog: xAI's 2T-Parameter Grok Model Nears Completion

    xAI's next 2T-parameter Grok model is expected to finish training next week, aiming to surpass Kimi K3 with better speed and token efficiency.

    Source

  3. Jul 01

    2026

    model

    TestingCatalog: Google testing new Gemini Flash checkpoint on LM Arena

    Google is testing a new Gemini Flash checkpoint on LM Arena that may release as Gemini 3.6 Flash or Gemini 4 Flash.

    Source

June

2026

39 events

  1. Jun 30

    2026

    model

    Claude Devs: Claude Sonnet 5 launches with 1M context window

    Anthropic releases Claude Sonnet 5, offering top-tier coding and tool-use performance with a 1 million token context window at Sonnet pricing.

    Source

  2. Jun 30

    2026

    model

    Claude Devs: Claude Sonnet 5 launches with 1M context window

    Claude Sonnet 5 is now available with top-tier coding and tool use performance, Sonnet pricing, and a 1 million token context window across the Claude Platform.

    Source

  3. Jun 30

    2026

    model

    Claude: Anthropic launches Claude Sonnet 5, its most agentic model yet

    Claude Sonnet 5 can make plans, use browsers and terminals, and run autonomously, matching capabilities previously only found in larger, costlier models.

    Source

  4. Jun 30

    2026

    model

    Claude: Claude Sonnet 5 Debuts as Most Agentic Model Yet

    Anthropic launches Claude Sonnet 5 with advanced autonomous agent capabilities, including tool use and planning.

    Source

  5. Jun 30

    2026

    model

    Claude: Claude Sonnet 5 launches with advanced agentic capabilities

    Claude Sonnet 5 can make plans, use browsers and terminals, and run autonomously at levels previously requiring larger and more expensive models.

    Source

  6. Jun 26

    2026

    model

    Sam Altman: OpenAI updates ChatGPT's 5.5 instant model

    Sam Altman announces that OpenAI updated the 5.5 instant model used in ChatGPT this week.

    Source

  7. Jun 26

    2026

    model

    Sam Altman: OpenAI launches Sol and Terra models in limited US preview

    Altman introduces Sol and Terra in the GPT-5.6 family but notes a US-government-requested limited preview instead of open access.

    Source

  8. Jun 26

    2026

    model

    Sam Altman: OpenAI launches Sol & Terra models in limited US preview

    Altman introduces GPT-5.6 family models Sol and Terra but reveals the US government requested a limited preview instead of open access, sparking a discussion on iterative deployment and regulatory cooperation.

    Source

  9. Jun 26

    2026

    model

    Sam Altman: OpenAI launches Sol and Terra models in limited preview

    Sam Altman announces new AI models Sol and Terra, but says the US government requested a limited preview launch instead of open access.

    Source

  10. Jun 11

    2026

    model

    Demis Hassabis: DiffusionGemma launches with 4x speed boost

    Demis Hassabis introduces DiffusionGemma, a new text diffusion model that is four times faster than other Gemma 4 variants.

    Source

  11. Jun 11

    2026

    model

    Demis Hassabis: DiffusionGemma text diffusion is 4x faster

    Celebrating DiffusionGemma, a new text diffusion model that runs four times faster than other Gemma 4 variants.

    Source

  12. Jun 11

    2026

    model

    Demis Hassabis: DiffusionGemma launches: 4x faster text diffusion

    Demis Hassabis celebrates the launch of DiffusionGemma, a new text diffusion model that is four times faster than other Gemma 4 variants.

    Source

  13. Jun 09

    2026

    model

    Andrej Karpathy: Karpathy hails Claude Fable 5 as major leap in AI-assisted coding

    Andrej Karpathy reviews Claude Fable 5, calling it a qualitative step change that enables ambitious coding tasks and boosts software demand via Jevons paradox.

    Source

  14. Jun 09

    2026

    model

    Andrej Karpathy: Karpathy: Claude Fable 5 is a major step forward for AI

    Karpathy reviews Claude Fable 5 as a major qualitative leap for difficult problem-solving, praising its benchmarks while noting overly cautious safeguards at launch.

    Source

  15. Jun 09

    2026

    model

    Claude: Anthropic launches Claude Fable 5, its most capable model yet

    Anthropic introduced Claude Fable 5, a Mythos-class model the company says is its safest and most capable generally available AI to date.

    Source

  16. Jun 04

    2026

    model

    TestingCatalog: NVIDIA drops Nemotron 3 Ultra on Hugging Face with 5× speed boost

    NVIDIA released Nemotron 3 Ultra, a 550B-parameter frontier LLM on Hugging Face promising 5× faster inference and 30% lower costs compared to other open models.

    Source

  17. Jun 04

    2026

    model

    TestingCatalog: Anthropic's claude-oceanus-v1-p surfaces for Red Teams

    A new Anthropic model called claude-oceanus-v1-p has been spotted available to Red Teams, hinting at an upcoming Mythos model release.

    Source

  18. Jun 04

    2026

    model

    TestingCatalog: Anthropic's Claude Oceanus v1-p surfaces for Red Teams

    Anthropic has made "claude-oceanus-v1-p" available to Red Teams, hinting at an upcoming release of newer Mythos models.

    Source

  19. Jun 04

    2026

    model

    Min Choi: Link to Google blog post on Gemma 4 12B launch

    Min Choi shared the official Google blog post with more details about the new Gemma 4 12B multimodal model.

    Source

  20. Jun 04

    2026

    model

    Min Choi: Gemma 4 12B skips separate encoders for direct multimodal processing

    Google's Gemma 4 12B processes vision and audio directly through its LLM backbone, cutting latency and memory to run locally on 16GB VRAM.

    Source

  21. Jun 04

    2026

    model

    Min Choi: Google drops open-source Gemma 4 12B multimodal model

    Google released Gemma 4 12B, an Apache 2.0 multimodal AI model that runs locally on laptops with vision, audio, reasoning, and agent capabilities.

    Source

  22. Jun 04

    2026

    model

    Min Choi: Google releases Gemma 4 12B open-source multimodal model

    Google launched Gemma 4 12B, an Apache 2.0 multimodal AI that runs locally on laptops without heavy encoders, supporting vision, audio, and agents.

    Source

  23. Jun 04

    2026

    model

    Min Choi: Microsoft AI launches 7 new MAI models toward superintelligence

    Microsoft AI launched seven new MAI models spanning reasoning, coding, image, voice, and transcription, describing the effort as a "hill-climbing machine" toward superintelligence.

    Source

  24. Jun 04

    2026

    model

    Min Choi: Microsoft AI launches 7 new MAI models toward superintelligence

    Microsoft AI released seven new MAI models spanning reasoning, coding, image, voice, and transcription as part of its broader push toward superintelligence.

    Source

  25. Jun 04

    2026

    model

    Min Choi: Microsoft AI launches 7 MAI models, teases superintelligence path

    Microsoft AI released seven new MAI models spanning reasoning, coding, image, voice, and transcription, framing them as steps toward superintelligence.

    Source

  26. Jun 04

    2026

    model

    Min Choi: Microsoft AI launches 7 new MAI models

    Microsoft AI released seven new MAI models spanning reasoning, coding, image, voice, and transcription, framed as a step toward superintelligence.

    Source

  27. Jun 03

    2026

    model

    TestingCatalog: Reve 2.0 launches with code-based editable image generation

    Reve 2.0 has launched, ranking second in text-to-image generation by representing images as editable code with precise layout control and segmentation.

    Source

  28. Jun 03

    2026

    model

    TestingCatalog: Reve 2.0 debuts as runner-up in text-to-image arena with code-based editing

    Reve 2.0 launched, ranking second in text-to-image benchmarks and introducing code-represented, segmented images for precise layout control.

    Source

  29. Jun 03

    2026

    model

    Demis Hassabis: Gemma 4 12B model released as downloads top 150 million

    Demis Hassabis celebrates Gemma 4 reaching over 150 million downloads and announces the release of the new Gemma 4 12B model, which is powerful yet compact enough to run locally on a laptop with 16GB VRAM under an Apache

    Source

  30. Jun 03

    2026

    model

    Demis Hassabis: Gemma 4 surpasses 150M downloads as new 12B model launches

    Demis Hassabis celebrates Gemma 4 reaching over 150 million downloads and announces the release of the new Gemma 4 12B model, which is powerful yet compact enough to run locally on a laptop with 16GB VRAM under an Apache

    Source

  31. Jun 03

    2026

    model

    TestingCatalog: Ideogram 4.0 launches as new open SOTA image model with 2K and transparency

    Ideogram released version 4.0, an open image generation model ranking highly on LM Arena with 2K resolution, accurate text, and background transparency.

    Source

  32. Jun 03

    2026

    model

    TestingCatalog: Ideogram 4.0 debuts as SOTA open image generation model

    Ideogram launched version 4.0, ranking 8th on LM Arena and 5th on Design Arena, with native 2K resolution, accurate text rendering, and background transparency.

    Source

  33. Jun 03

    2026

    model

    TestingCatalog: Ideogram 4.0 debuts as new open image generation model

    Ideogram launched version 4.0, ranking 8th on LM Arena and 5th on Design Arena with features like 2K resolution and background transparency.

    Source

  34. Jun 03

    2026

    model

    TestingCatalog: Ideogram 4.0 debuts as new open SOTA image generation model

    Ideogram released version 4.0, an open image generation model ranking eighth on LM Arena with native 2K resolution and precise layout control.

    Source

  35. Jun 03

    2026

    model

    TestingCatalog: Ideogram 4.0 debuts as SOTA open image generation model

    Ideogram released version 4.0 of its open image generation model, which ranks eighth on LM Arena and supports 2K resolution and precise text rendering.

    Source

  36. Jun 03

    2026

    model

    TestingCatalog: Ideogram 4.0 debuts as new open image generation SOTA

    Ideogram released version 4.0, ranking 8th on LM Arena and 5th on Design Arena, featuring 2K resolution, text rendering, and layout control.

    Source

  37. Jun 03

    2026

    model

    TestingCatalog: Google releases Gemma 4 12B encoder-free multimodal model

    Google released Gemma 4 12B on Hugging Face under Apache 2.0, an encoder-free multimodal model supporting text, audio, image, and video for local devices.

    Source

  38. Jun 03

    2026

    model

    TestingCatalog: Google releases Gemma 4 12B open-source multimodal model

    Google's new Gemma 4 12B model is available on Huggingface under Apache 2.0, offering encoder-free multimodal capabilities ideal for local consumer device deployment.

    Source

  39. Jun 03

    2026

    model

    TestingCatalog: Google releases Gemma 4 12B multimodal model on Hugging Face

    Google launched Gemma 4 12B under Apache 2.0 on Hugging Face, bringing encoder-free multimodal capabilities for local consumer device deployment.

    Source

May

2026

9 events

  1. May 28

    2026

    model

    Min Choi: Claude Opus 4.8 drops with major coding and agent upgrades

    Anthropic released Claude Opus 4.8 featuring improved coding, enhanced agent capabilities, effort control, and dynamic workflows with hundreds of subagents, plus Mythos-class models coming soon.

    Source

  2. May 28

    2026

    model

    Claude: Anthropic launches Claude Opus 4.8 with sharper judgment

    Claude announced Opus 4.8, featuring sharper judgment, improved honesty about its progress, and longer independent work at the same price.

    Source

  3. May 28

    2026

    model

    Claude: Anthropic launches Claude Opus 4.8 with sharper judgment

    Claude Opus 4.8 is now available, featuring improved judgment, greater self-assessment honesty, and extended independent operation at unchanged pricing.

    Source

  4. May 28

    2026

    model

    Claude: Anthropic launches Claude Opus 4.8 with sharper judgment

    Claude Opus 4.8 launches with improved judgment, greater honesty about its capabilities, and extended independent work sessions at unchanged pricing.

    Source

  5. May 20

    2026

    model

    Demis Hassabis: More info on Gemini 3.5 Flash model released

    Demis Hassabis shares a link to Google's blog post with more information about the Gemini 3.5 Flash model.

    Source

  6. May 20

    2026

    model

    Demis Hassabis: Hassabis shares details on new Gemini 3.5 Flash model

    Demis Hassabis shares a link to Google's blog post with more information on the Gemini 3.5 Flash model.

    Source

  7. May 20

    2026

    model

    Demis Hassabis: Gemini 3.5 Flash launches with 4x speed and half the cost

    Demis Hassabis introduces Gemini 3.5 Flash, highlighting superior coding and agentic performance over 3.1 Pro, 4x speed gains, and lower costs, now available in Antigravity and Gemini App.

    Source

  8. May 20

    2026

    model

    Demis Hassabis: Gemini 3.5 Flash outperforms 3.1 Pro at lower cost

    Demis Hassabis announces Gemini 3.5 Flash, highlighting superior coding performance, 4x speed, and half the cost of frontier models.

    Source

  9. May 20

    2026

    model

    Demis Hassabis: Gemini 3.5 Flash outperforms 3.1 Pro, 4x faster, cheaper

    Hassabis touts Gemini 3.5 Flash for outperforming 3.1 Pro on coding and agentic tasks while being four times faster and often less than half the cost, with a Pro version coming soon.

    Source

July

2025

2 events

  1. Jul 17

    2025

    model

    Min Choi: Grok 4.5 outperforms larger models at a fraction of the cost

    Min Choi highlights Grok 4.5's efficiency advantage over Kimi K3, noting it achieves better results with fewer parameters and lower cost per task.

    Source

  2. Jul 11

    2025

    model

    Min Choi: Grok 4 sparks a building frenzy with its powerful capabilities

    Min Choi declares Grok 4's capabilities are driving a wave of creative development, showcasing ten examples of what people are building with the model.

    Source

May

2025

2 events

  1. May 22

    2025

    model

    Min Choi: Google Veo 3's hyper-realistic AI video stuns the internet

    Min Choi showcases ten jaw-dropping examples of Google Veo 3's AI-generated video realism, including fabricated street interviews that look indistinguishable from real footage.

    Source

  2. May 22

    2025

    model

    Claude Opus 4 introduces extended thinking

    First mainstream model with developer-tunable 'thinking budget'. Establishes the pattern other labs subsequently follow.

January

2025

1 event

  1. Jan 20

    2025

    model

    DeepSeek V3 + R1 release — the Sputnik moment

    A Chinese lab releases an open-weights reasoning model at frontier quality, trained for roughly $5M. Triggers an immediate global reassessment of the cost frontier.

August

2024

1 event

  1. Aug 11

    2024

    model

    Min Choi: Flux AI's photorealistic images shock the internet

    Min Choi highlights ten stunning examples of Flux AI's image generation realism, emphasizing that the outputs are entirely AI-generated.

    Source

June

2024

1 event

  1. Jun 27

    2024

    model

    Min Choi: Kling AI brings internet memes to life in stunning video

    Min Choi showcases eight examples of Kling AI generating viral meme-inspired video content, highlighting its impressive creative capabilities.

    Source

May

2024

1 event

  1. May 13

    2024

    model

    Min Choi: OpenAI's ChatGPT-4o launch is a game-changer for AI assistants

    Min Choi highlights ten compelling demonstrations of ChatGPT-4o's real-time visual and conversational capabilities, declaring it will transform the AI assistant landscape.

    Source

April

2024

1 event

  1. Apr 18

    2024

    model

    Meta releases Llama 3

    8B and 70B open-weights models. Llama 3 70B is the first open model to clearly approach closed-frontier quality on benchmarks.

March

2024

1 event

  1. Mar 04

    2024

    model

    Anthropic releases Claude 3 family — Opus tops the boards

    Opus becomes the first non-OpenAI model to clearly top frontier benchmarks. Mid-tier Sonnet and small Haiku launch alongside.

February

2024

1 event

  1. Feb 15

    2024

    model

    OpenAI shows Sora — minute-long generated video

    Quality leap so large that many doubted it was real. Public release follows months later.

December

2023

1 event

  1. Dec 08

    2023

    model

    Mistral releases Mixtral 8x7B (MoE) with open weights

    First widely-used open-weights MoE model. Sets the template for the next two years of European open-weights frontier work.

March

2023

2 events

  1. Mar 14

    2023

    model

    OpenAI releases GPT-4

    First multimodal frontier model. Doubles down on alignment work via RLHF and constitutional methods. Capability jumps over GPT-3.5 are massive.

  2. Mar 14

    2023

    model

    Anthropic introduces Claude

    Same day as GPT-4. Anthropic positions Claude on safety-by-design via Constitutional AI — the first serious alternative to OpenAI as a frontier lab.

August

2022

1 event

  1. Aug 22

    2022

    model

    Stable Diffusion v1 released with open weights

    The image-generation revolution becomes open-source overnight. Spawns a thousand forks and the modern AI image ecosystem.

May

2020

1 event

  1. May 28

    2020

    model

    GPT-3 paper and API launch

    175B parameters. Few-shot prompting becomes a thing. The first model to feel categorically different.

February

2019

1 event

  1. Feb 14

    2019

    model

    OpenAI releases GPT-2 (cautiously)

    OpenAI initially withholds the full 1.5B model citing misuse concerns — the first 'too dangerous to release' moment.

October

2018

1 event

  1. Oct 11

    2018

    model

    Google releases BERT

    First widely-adopted Transformer encoder. Sweeps NLP benchmarks; becomes the default building block for the next two years.