Google announced that its Antigravity SDK now supports local AI model workflows, with initial support for Gemma 4 26B A4B running through Google AI Edge's LiteRT runtime. The company recommends a mach…
Google released its Gemini 3.8 text-to-speech models, expanding the Gemini audio family that also includes Gemini 3.8 Live for real-time voice interactions and Gemini 3.8 Flash for coding tasks. The m…
Google has begun internal testing of early Gemini 4 Pro checkpoints under the code name "argon," according to a leak from Lentils, with developer Bee posting a demo on X of a monographic website the m…
A developer built Spud, a local-first push-to-talk dictation tool, by pairing with Gemini 3.8 Flash through Google's agy CLI, integrating it with the Hyprland Wayland compositor on Arch-based Endeavou…
Google limited Gemini 2.5 model access in its API on September 18, 2026 to users who have actively used the models in the past, stating the 2.5 models are not deprecated and will continue to be served…
Google fell to 10th place among AI labs on version 4.3.2 of the Artificial Analysis Intelligence Index, with its best-ranked model, Gemini 3.8 Flash, sitting 14th on the model-level chart at a score o…
Google's Android Bench 2.0 benchmark now measures long-horizon Android development tasks, with GPT 6 Astra (codex) leading at a 28.0% pass rate across 30 tasks, followed by Claude Fable 5 (claude-code…
The Rails Foundation reported on September 21st that GPT-6 Astra kept first place in its feature-development benchmark and raised its solve rate from 35% at medium reasoning effort to 53.3% at maximum…
Chinese AI lab StepFun released Step 5 Preview, a flagship agentic model that scored 44 on the Artificial Analysis Intelligence Index, beating Google's Gemini 3.8 Flash (high) at 41 while costing $0.7…
Gemini 3.8 Flash identified spider species correctly on 997 of 2,000 photos, a 49.85% accuracy rate that was the highest among nine AI models tested in a benchmark built from research-grade iNaturalis…
Developer lukstei released slop-grader, a CLI tool that grades text files against custom rulesets using the Jev System One model from typesafe.ai. The tool batches lines into groups of 255 and evaluat…
Five frontier-class AI models shipped between September 1 and September 10, 2026 — Anthropic's Claude Fable 5.1, Google's Gemini 3.8 Flash, Meta's Muse Spark 1.3, OpenAI's GPT-6 Astra, and DeepSeek's …
Qwen released Qwen3.8-Omni-Flash, its first multimodal model built for AI agents, which processes audio and video together and independently uses tools to edit vlogs, translate clips, or summarize mov…
Google shipped a preview of Managed Agents in the Gemini API in September 2026, introducing the antigravity-preview-09-2026 environment and upgrading the default model from Gemini 3.5 Flash to Gemini …
Google rebuilt its CC AI agent as a shared household assistant that coordinates calendars, tasks, documents, forms, and daily briefings for up to six adults, available as an early Google Labs experime…
Alibaba's Qwen team released Qwen3.8-Omni-Flash, which it calls its first omni-modal model built around agentic capabilities, accepting text, images, audio, and video and returning text. The model is …
A benchmark of frontier models on production-grade CAD assembly design found none reached the 60% pass threshold as of September 16, 2026, with GPT 6 (Astra, Codex at max) scoring highest at 74% ± 8% …
Several X users reported that Google's Gemini 4 Pro is being tested on Arena under the model name 'Gemini 3.8 Flash', with one user citing a 256k output ceiling as evidence the model is Gemini 4 Pro. …
A developer's side-by-side test found that Google's Gemini 3.8 Flash cloud model produced a faster Rust parsing function than the local Gemma 4 model (gemma-4-26B_q4_0-it.gguf), cutting starts_with ca…
Independent evaluator Vals found that Google's Gemini 3.8 Flash scored 71.7% on BioMysteryBench's human-solvable tasks and 21.6% on its hard tasks in production runs, versus the 88.8% and 56.5% Google…