Anthropic shipped 6 of the 14 frontier models since July. OpenAI ships more often, mostly below the frontier

Since July 1, 78 AI models from 37 labs have gone live on a public API, by ArtificialWatch's own release log. Fourteen of them were frontier releases, and six of those 14 came from Anthropic: Fable 5 (July 1), Claude Opus 5 (July 24), Fable 5.1 and Mythos 5.1 (both September 1), Claude Opus 5.5 (September 22) and Claude Sonnet 5.5 (September 28). No other lab shipped more than three.

HOW OFTEN EACH LAB SHIPS

The median gap between release days from July 1 to November 3, all models counted (two models on the same day count as one release day):

- OpenAI: about every 10 days (8 releases on 7 days)

- Google: about every 13 days (7 releases on 6 days)

- Meta: about every 14 days (8 releases on 7 days)

- Anthropic: about every 21 days (7 releases on 6 days)

- xAI: about every 22 days (5 releases on 5 days)

FRONTIER RELEASES COME LESS OFTEN

Counting frontier models only, Anthropic's median gap is 22 days. OpenAI's is 41: ChatGPT-5.6 on July 9, GPT-6 Astra on September 3 and GPT-6.1 Sol on September 29. Google, Meta and xAI shipped one frontier model each in the window: Gemini 4 (October 13), Muse Wonder (October 12) and Grok 4.7 (September 21). The other two were Alibaba's Qwen 3.8 Max (August 1) and Moonshot's Kimi K3.1 Pro (October 23).

THE MEDIAN HIDES THE BURSTS

OpenAI went 56 days without a release between ChatGPT-5.6 and GPT-6 Astra, then shipped on five days in four weeks: GPT-6 Astra, GPT-Image-2.5 (September 8), GPT-Live-1 (September 10), GPT-6 Sol and GPT-6 Luna (September 22) and GPT-6.1 Sol (September 29). Its latest, GPT-6.1 Terra, went live on October 26. Across all labs September was the busiest month, with 27 releases, against 15 in July, 19 in August and 17 in October. A new model went live on 53 of the 126 days.

WHERE THINGS STAND ON NOVEMBER 3

Anthropic's latest frontier release, Claude Sonnet 5.5, went live 36 days ago, longer than its 22-day median. OpenAI's, GPT-6.1 Sol, went live 35 days ago, against a 41-day median. Past gaps do not schedule future launches; they only say what has been normal since July.

WHAT THESE NUMBERS ARE

Each date is the day ArtificialWatch verified the model live on a public API, not the day it was announced. "Frontier" is our own flag for a lab's top-tier model. The window starts July 1 because our log before then is incomplete. The counts include image, voice and transcription models and cover the models we track, not every fine-tune or open checkpoint. A median of five or six gaps is a small sample, so one launch can move it by days. Each lab's live figures are on its release page at artificialwatch.com/releases/lab, for example /releases/lab/openai and /releases/lab/anthropic.