AI Daily · Oct 15, 2026: Google launches Gemini 4 Argon for AI Pro and Ultra subscribers
Google's frontier model Gemini 4 Argon rolls out globally to AI Pro and Ultra subscribers, seeing 3x Claude's launch-day adoption on Cline with 30% fewer tokens; StepFun open-sources its 600B MoE Step 5 Preview with 1M-token context.
Showing the latest briefing (Oct 15, 2026). Every day is generated the next morning.
Pick a date
-
Google launches Gemini 4 Argon for AI Pro and Ultra subscribers
Gemini 4 Argon is now available globally to Google AI Pro and Ultra subscribers via the Gemini app and web. Google also published a conversation with the model's builders.
-
StepFun releases Step 5 Preview open weights with 1M-token context
Step 5 Preview is now open weights (BF16 and FP8) on Hugging Face and ModelScope: a sparse MoE with 600B total and 27B active parameters per token, 1M-token context, native text and vision input, built for long-horizon agentic work. It supports OpenAI-compatible serving via vLLM or SGLang, offline Transformers inference, and Day-0 NVIDIA support; Step Code was updated to v0.1.4.
-
Grok Voice 3.0 launches with sub-second responses and $0.08/min API pricing
SpaceXAI released Grok Voice 3.0, a speech-to-speech model that responds in under a second with 40% lower latency, resolves 70% of Starlink support calls on its own, and is available today in the API and Voice Agent Builder at $0.08/min.
-
NVIDIA Nemotron 3.5 Super VL goes live on OpenRouter
NVIDIA's open 120B-A12B multimodal Nemotron 3.5 Super VL is available on OpenRouter with 1M context, handling text, images, documents, GUIs and video; it uses a hybrid Mamba-Transformer MoE with multi-token prediction.
-
Black Forest Labs releases FLUX 3 Image Klein with open weights
FLUX 3 Image Klein is an open-weights image model with 5.6B parameters that runs on a single consumer GPU, supporting native 4K, bounding boxes and pixel-perfect editing.
-
DeepSeek API outage hit service for nearly five hours
A DeepSeek engineer said the API service failed between 8:11 and 13:04 Beijing time on October 15, 2026, and that an investigation and rectification followed.
-
Gemini 4 Argon sees 3x launch-day adoption on Cline, uses 30% fewer tokens
On Cline, Gemini 4 Argon had the best launch-day user share of any frontier model since June, with ~3x as many users trying it on day one as Claude Opus 5.5 or Sonnet 5.5. It also uses 30% fewer tokens on average than Claude models.
-
OpenAI introduces ChatGPT Space for team collaboration at DevDay
At DevDay OpenAI launched ChatGPT Space, where teams create living documents with ChatGPT, turn messy conversations into organized plans, and invite teammates to edit or comment.
-
Command Code GOAT plan gives $100 usage for $10
Command Code's $10 GOAT plan now provides $100 of usage; Ling 3.1 Flash, GLM 5.3 Flash and DeepSeek V4.1 Flash each get $100 usage permanently, and higher plans include GOAT plus extra usage.
-
ElevenLabs models 50% off on OpenRouter through October 19
OpenRouter reports ElevenLabs took nearly a third of text-to-speech spend yesterday after 7 days live, and is offering all ElevenLabs models at 50% off exclusively on OpenRouter through 10/19.