What changed in AI.

Claude enters Google Workspace public beta

Anthropic introduced a Google Docs, Sheets and Slides sidebar in public beta for paid plans.

Launched EmbeddingGemma 2 on October 6, 2026, a 740M-parameter multimodal embedding model for on-device search…

Launched EmbeddingGemma 2 on October 6, 2026, a 740M-parameter multimodal embedding model for on-device search under Apache 2.0.

Launched Mistral Large 4 in public preview on October 6, 2026 through a preview API

Launched Mistral Large 4 in public preview on October 6, 2026 through a preview API; open weights are planned for later October.

OpenAI publishes AI mathematics manuscripts

OpenAI released hundreds of manuscripts from an internal model; not all results are formalized or fully verified, and the catalogue continues to change.

Opened the Decisions API in public beta on October 6, 2026.

TII launches Falcon-Emirati, Falcon-ASR and Falcon-OCR-Arabic

TII launched three Arabic-focused models: a 7B Emirati Arabic language model, a 1.6B multilingual speech recognition model, and an Arabic OCR/document understanding model.

Nolla launches supervised Utah acne pilot

Launched a supervised Utah acne-prescription pilot on October 5, 2026 for adults, limited to preapproved topical protocols. Initial prescriptions require two-physician approval; later stages need state authorization.

Launched North 2 on October 5, 2026, an enterprise agent platform with memory, libraries, skills and flexible…

Launched North 2 on October 5, 2026, an enterprise agent platform with memory, libraries, skills and flexible cloud, on-premises and air-gapped deployment.

Introduced Beam on October 5, 2026 in a select-user research preview while completing red-teaming

Introduced Beam on October 5, 2026 in a select-user research preview while completing red-teaming; weights and developer artifacts are planned for later release.

Released Together Link on October 5, 2026, an open-source connector routing supported coding clients to Together…

Released Together Link on October 5, 2026, an open-source connector routing supported coding clients to Together models through user-provided API keys.

Cactus Compute releases Whistle speech model

Cactus Compute released Whistle, a 16.9 MB open speech-recognition model for seven languages with CPU-only operation.

Partnered with Smallest.ai to run Lightning V2 text-to-speech on Galaxy Blackhole servers for on-premises…

Partnered with Smallest.ai to run Lightning V2 text-to-speech on Galaxy Blackhole servers for on-premises enterprise voice AI, announced October 2, 2026.

Introduced branch-based restores for Lakebase Postgres on October 1, 2026, using immutable history rather than…

Introduced branch-based restores for Lakebase Postgres on October 1, 2026, using immutable history rather than database copying.

Made its latest real-time TTS model the default for API and Studio on October 1, 2026, exiting beta

Made its latest real-time TTS model the default for API and Studio on October 1, 2026, exiting beta; the company reports about 50ms to first audio chunk.

Launched Video Editor in Manus 2.0 on October 1, 2026 for macOS and Windows, with separate editable video,…

Launched Video Editor in Manus 2.0 on October 1, 2026 for macOS and Windows, with separate editable video, captions, graphics and audio tracks.