AI Daily · Nov 11, 2026: Google launches Gemini 4 Carbon for long-horizon agentic tasks
Google's strongest model, Gemini 4 Carbon, is rolling out to Ultra subscribers and the API for long-horizon agentic work; OpenRouter, Cursor and LiteLLM offer 50% introductory pricing, with Cursor's rates doubling December 1.
Pick a date
-
Google launches Gemini 4 Carbon for long-horizon agentic tasks
Google released Gemini 4 Carbon, described as its strongest model, aimed at sustained agentic workflows in knowledge work, web development and cyber defense; Gemini 4 Argon remains the default for most software engineering work.
-
Cursor offers Gemini 4 Carbon at 50% off until November 30
In Cursor, Gemini 4 Carbon costs $2 per million input and $10 per million output tokens through November 30, then rises to $4 and $20 per million on December 1.
-
Gemini 4 Carbon goes live on OpenRouter at 50% introductory pricing
Gemini 4 Carbon is available on OpenRouter via Google Cloud, with introductory pricing of $2 per million input tokens and $1 per million output tokens, 50% off through Nov 30. Google reports gains over Gemini 4 Argon on vulnerability exploitation, agentic terminal tasks and long-context reasoning.
-
Requesty lists Gemini 4 Carbon at $1 input, $5 output per million tokens
AI gateway Requesty now serves Gemini 4 Carbon under the model name gemini-4-carbon, priced at $1 per million input tokens and $5 per million output tokens, aimed at multi-step coding and agent workloads.
-
Moonshot releases Kimi K3.1 Pro, demoed on Cerebras hardware
Moonshot AI released Kimi K3.1 Pro; a Cerebras demo used it to complete an online shopping task from a single prompt in 11.5 seconds.
-
Alibaba previews Qwen4-Max on Token Plan and QwenWork
Qwen announced Qwen4 with stronger coding and agentic capabilities; Qwen4-Max is in preview on Alibaba's Token Plan and offers early access via QwenWork and Qode.
-
Qoder ships Qwen4-Max-Preview to all plans with 80% discount
Qoder made Qwen4-Max-Preview available day-one across every product and plan including Teams and Enterprise, with a limited-time 80% discount at 0.1x credits; the build updates daily and is still a preview.
-
Stripe acquires OpenRouter, used by over 10 million developers
OpenRouter, which offers 400+ models including 50+ free and serves more than 10 million developers and companies, is now part of Stripe; the companies frame it as infrastructure for token routing and cost control.
-
LightOnOCR-3-0.8B processes 14,483 pages at 7.7 pages per second
A user ran LightOnOCR-3-0.8B from LightOn on one RTX PRO 6000 via Hugging Face Jobs with vLLM, OCRing 14,483 pages of fossil fuel industry documents at about 7.7 pages/s once running, for $1.67 total or roughly $0.12 per 1,000 pages.
-
OpenRouter reports GLM 5.3 Flash TTFT down 40% after routing change
OpenRouter says its new routing algorithm, which sends more traffic to providers with good inference performance, cut GLM 5.3 Flash time-to-first-token by 40%, raised throughput 10% and lowered cost per 1M tokens by 17% over the past month.