# AI model releases, 2026

> 17 frontier AI model releases tracked across 2026, month by month, with what each one made automatable. Updated through August 2026.

## Full timeline

| Date | Model | Lab | Capability |
| --- | --- | --- | --- |
| 4 August 2026 | Jalapeño inference chip | OpenAI and Broadcom | Custom inference silicon, claimed ~50% lower cost per token |
| 29 July 2026 | Claude Opus 5 | Anthropic | Near-Fable-5 intelligence at roughly half the price |
| 29 July 2026 | Gemini 3.6 Flash and 3.5 Flash-Lite | Google | Google's most efficient agentic models to date |
| 14 July 2026 | GPT-5.6 (Sol, Terra, Luna) | OpenAI | Tiered autonomous agents spanning cost and capability |
| 9 June 2026 | Claude Fable 5 and Mythos 5 | Anthropic | Adaptive thinking - reasoning effort scaled to the task |
| 24 April 2026 | DeepSeek V4 | DeepSeek | Near-frontier performance, released open source |
| 23 April 2026 | GPT-5.5 | OpenAI | Autonomous multi-step research and agentic coding |
| 17 April 2026 | Claude Design | Anthropic | Generative prototyping, built with Canva |
| 16 April 2026 | Claude Opus 4.7 | Anthropic | Elite software engineering with sustained multi-file consistency |
| 14 April 2026 | Gemma 4 | Google | Open-weight models capable of running locally |
| 7 April 2026 | Claude Mythos | Anthropic | Cybersecurity-specialised frontier model |
| 2 April 2026 | Claude Code | Anthropic | Agentic engineering rather than code assistance |
| 24 March 2026 | Claude Cowork | Anthropic | Agentic memory layer carrying context across a team |
| 5 March 2026 | GPT-5.4 | OpenAI | Flagship 'Thinking' mode - deliberate multi-step reasoning |
| 19 February 2026 | Gemini 3.1 Pro | Google | Multimodal reasoning across image, code and text |
| 17 February 2026 | Claude Sonnet 4.6 | Anthropic | Versatile coding at materially lower cost |
| 5 February 2026 | Claude Opus 4.6 | Anthropic | Advanced coding and agentic task execution |

## By month

- **August 2026** (1 release): No new flagship - the story moved to inference economics, and OpenAI's custom Jalapeño chip put a number on how much cheaper automation is about to get. — https://jobsecuritymeter.com/ai-model-releases/august-2026
- **July 2026** (3 releases): Three frontier releases: GPT-5.6's Sol, Terra and Luna tiers, Claude Opus 5, and Gemini 3.6 Flash - the month efficiency became the headline feature. — https://jobsecuritymeter.com/ai-model-releases/july-2026
- **June 2026** (1 release): Anthropic opened the Claude 5 generation with Fable 5 and Mythos 5, introducing adaptive thinking effort. — https://jobsecuritymeter.com/ai-model-releases/june-2026
- **May 2026** (no release): No new flagship shipped in May - the frontier held at Claude Opus 4.7, GPT-5.5 and Gemini 3.1 Pro while the industry absorbed April. — https://jobsecuritymeter.com/ai-model-releases/may-2026
- **April 2026** (7 releases): The April Sprint: seven releases in a month, with Claude Opus 4.7, GPT-5.5 and DeepSeek V4 landing inside eight days. — https://jobsecuritymeter.com/ai-model-releases/april-2026
- **March 2026** (2 releases): OpenAI answered with GPT-5.4 and its 'Thinking' mode, extending frontier reasoning into finance, law and strategy work. — https://jobsecuritymeter.com/ai-model-releases/march-2026
- **February 2026** (3 releases): Three frontier models in fourteen days - Claude Opus 4.6, Claude Sonnet 4.6 and Gemini 3.1 Pro - opened the fastest release year on record. — https://jobsecuritymeter.com/ai-model-releases/february-2026

---

**Source:** Job Security Meter — https://jobsecuritymeter.com/ai-model-releases
**Human-readable version:** https://jobsecuritymeter.com/ai-model-releases
**Last updated:** August 2026
**Attribution:** Free to quote and cite with attribution to Job Security Meter.
**Full site specification for language models:** https://jobsecuritymeter.com/llms-full.txt