Anthropic led model release activity in Scoopkit's event database with 36 events, followed by Alibaba at 24, Google at 24, OpenAI at 24, Z.ai at 11. This comes from 288 total MODEL_RELEASE category events counted directly, not estimated.

This is a live count, not a curated list. It reflects whatever Scoopkit's ingestion pipeline picked up and classified as a model release in the window checked.

Methodology

Counted from Scoopkit's public /v1/events endpoint, filtered to category=MODEL_RELEASE, on 2026-10-12. The free API tier keeps a rolling 45-day archive, so the window covers 2026-07-15 through 2026-10-12. Org attribution comes from automated entity extraction on the event, which means near-duplicate names, like "Google" and "Google DeepMind," get counted separately rather than merged. That's a real limitation of this count, not hidden below. This page refreshes automatically on a recurring basis, so the numbers above reflect whatever window was current on 2026-10-12, not a fixed snapshot.

Facts: releases by lab, last 45 days

RankLabModel release events
1Anthropic36
2Alibaba24
2Google24
2OpenAI24
5Z.ai11
5Moonshot AI11
7Google DeepMind9
8NVIDIA8
8OpenRouter8
10Meta7
11Qwen6
11Liquid AI6
13Microsoft5
13AWS5
13Artificial Analysis5
13DeepSeek5
13ByteDance5

163 more distinct names appeared with fewer events each (214 events total), not listed individually here. Mostly single mentions of individuals, policy bodies, or smaller teams picked up by automated entity extraction, not labs shipping models at any real cadence.

What kind of release these actually are

Of the 288 events, 180 were new model releases, 58 were open-weights releases, 24 were benchmark results, 23 were version updates, 1 was a benchmark saturation event, 1 was a license change, and 1 was a model deprecation.

By month

July 2026 had 42 of these events; August 2026 had 65 of these events; September 2026 had 128 of these events; October 2026 had 53 of these events. A month still in progress at the time of this check is a partial-month count, not a sign of an actual trend.

Query this same category yourself with a free API key, or see the full field reference at the API docs.

FAQ

How often does this get updated?

This snapshot is from 2026-10-12, refreshed automatically on a recurring basis rather than published once, since a 45-day rolling window changes every day it isn't updated.

Why do some lab names look duplicated?

Org names come from automated entity extraction, not a manually curated list. Near-duplicate strings like "Google" and "Google DeepMind" are extracted separately rather than merged into a guess about which one is more correct.

Where does this data come from?

Scoopkit's own event database, queried live through the public API, the same one described in how Scoopkit deduplicates AI news headlines.