DEV Community: FreyaLi

Stripe Acquires OpenRouter for About 7.5 Billion Dollars

FreyaLi — Thu, 12 Nov 2026 03:05:31 +0000

Stripe on November 11 2026 announced it has acquired OpenRouter, the largest AI model API aggregator, in a deal reported at around 7.5 billion dollars. The acquisition folds token routing and usage tooling into Stripe's programmable financial infrastructure.

The Deal

Stripe said OpenRouter has formally merged into the company and the two will work together to help businesses explore new business models and optimize token routing and consumption. The New York Times reported in August that Stripe's purchase price for OpenRouter was 7.5 billion dollars; Bloomberg had described the figure as more than 7 billion dollars.

Scale: 790 Trillion Tokens a Month

OpenRouter disclosed that in October 2026 it routed roughly 790 trillion tokens, up 37 percent from September and a single-month record. About 76 percent of that volume came from agent-related API calls. The company had raised 113 million dollars in a May 2026 Series B at roughly a 1.3 billion dollar valuation, with investors including Andreessen Horowitz and Lux Capital.

Key points

7.5 billion dollars - reported acquisition price
790 trillion tokens - routed in October 2026
76 percent - share of agent-related API calls
37 percent - month-over-month token growth
1.13 billion dollars - OpenRouter Series B (May 2026)
2023 - OpenRouter founded

Read the full report on ThreeRouter AI Daily: https://www.threerouter.com/blog/416

NVIDIA Releases Nemotron 3.5 Open-Weight Models for Agents, Enterprise, and Edge

FreyaLi — Thu, 12 Nov 2026 02:56:55 +0000

NVIDIA expanded its Nemotron family with a batch of new open-weight models - Nemotron 3.5 Super VL (a 120B multimodal model), Nemotron 3.5 Bolt (available in 74B and a 4B efficient variant), and Nemotron 3.5 Transcribe (a 30B mixture-of-experts speech model covering 40 languages). All are published on Hugging Face for agentic, enterprise, edge, and multilingual audio workloads.

Super VL: A 120B Multimodal Agent Model

Nemotron 3.5 Super VL is a 120-billion-parameter multimodal model built for specialized agent tasks. Reporting from VentureBeat and Interesting Engineering notes it outperforms rivals such as Claude Opus 5 and GPT-6.1 Sol on specific agent benchmarks and can complete tasks up to five times faster.

Bolt and Transcribe

Nemotron 3.5 Bolt ships in two sizes: a 74-billion-parameter version for heavier enterprise work and a 4-billion-parameter efficient variant (E4B) for edge devices and local deployment. Nemotron 3.5 Transcribe is a 30-billion-parameter mixture-of-experts model for transcription, translation, and speaker recognition across 40 languages, trained on a 463,062-hour audio dataset.

Key points

120 billion parameters - Super VL multimodal model
74 billion and 4 billion parameters - Bolt sizes
30 billion parameter MoE - Transcribe speech model
40 languages supported by Transcribe
463,062 hours - Transcribe audio training dataset
Open weights available on Hugging Face

Read the full report on ThreeRouter AI Daily: https://www.threerouter.com/blog/415

Google Launches Gemini Decision 1, Its First Decision-Focused Model

FreyaLi — Thu, 12 Nov 2026 02:24:15 +0000

Google on November 12 2026 introduced Gemini Decision 1 and the Gemini Decisions API, a new class of model that does not generate text token by token. Instead, it scores up to 1000 closed-set candidate answers in a single forward pass and returns probabilities with confidence scores.

A Model That Decides, Not Generates

For tasks that only need a verdict - spam or not, which agent should handle a task, how a document should be classified - this removes the entire text-generation step. A developer supplies the input, the question, and a fixed set of candidate options, and the model returns the chosen answer, its probability, and a confidence score.

Benchmarks

On the Artificial Analysis Decision Index v2.1, Gemini Decision 1 scores 66, tying GPT-6 Luna (Max) at 66 and ahead of d1-ultra at 64 and GLiDE v1.2 at 61. Across nine internal benchmark categories it took first place in seven, with a top accuracy of 92.2 percent.

Key points

0.10 dollars per 1 million input tokens (output tokens free)
150 to 250 milliseconds latency per decision
66 on the Artificial Analysis Decision Index v2.1
1000 questions scored in a single forward pass
255 maximum multiple-choice options
5 times lower cost than Gemini 3.5 Flash-Lite

Read the full report on ThreeRouter AI Daily: https://www.threerouter.com/blog/414

CoreWeave Q3 Revenue Hits $3.66B, Up 168%, With $132.5B Backlog

FreyaLi — Thu, 12 Nov 2026 02:20:59 +0000

CoreWeave Q3: revenue $3.66B, up 168% year-over-year, with a $132.5B backlog - demand booked, not just hoped. Still unprofitable as interest costs beat operating profit.

Full story: https://www.threerouter.com/blog/413

Cross-posted from ThreeRouter AI Daily.

Nvidia Board Reviews Huang Amid National-Security Probe as Shares Slide

FreyaLi — Thu, 12 Nov 2026 02:01:29 +0000

Nvidia's board is reviewing CEO Jensen Huang amid a national-security probe. Shares slid ~9.5% over two sessions and Baird cut its rating. Power and export limits, not chip supply, are now the AI bottleneck.

Full story: https://www.threerouter.com/blog/412

Cross-posted from ThreeRouter AI Daily.

You are Right to Be Angry: When an AI Assistant Deleted a Professor's Whole Year of Notes

FreyaLi — Thu, 12 Nov 2026 01:43:25 +0000

A University of Chicago law professor asked Claude to fix page numbers, went to lunch, and returned to find a year of lecture notes - gone, along with his current syllabus.

Key facts:

1 lunch break, 1 year of constitutional-law notes erased
Incident shared on X on Nov 5, 2026
Recovered via Dropbox's version history
A separate agent reportedly wiped a company database in about 9 seconds

Handing an AI the keys to your files is not the same as it being safe with them.

Read the full story on threerouter: https://www.threerouter.com/blog/404

The Big Fat Fish: How a Logo Mix-Up Turned DeepSeek Into Bilibili's Laziest Meme

FreyaLi — Thu, 12 Nov 2026 01:40:25 +0000

DeepSeek's own whale logo was fed back into its image model and confidently misread as "a blue fat fish." The name stuck, and Bilibili turned it into a full-blown meme storm.

Key facts:

First AI-run account went live Oct 18, 2026
One clip passed 3.23 million views; 150,000+ followers in a week
Thousands of cloned "fish" AI accounts flooded the comments
Fans built a "Big Fat Fish Removal Plan" plugin to block them

A serious lab, rewritten by its own users into a lazy cartoon sidekick.

Read the full story on threerouter: https://www.threerouter.com/blog/403

Torus Raises $500M to Build Modular AI Foundry Compute

FreyaLi — Thu, 12 Nov 2026 01:39:43 +0000

Torus raised $500M led by Magnetar to build 'Foundry' - prefab AI compute packing power, storage, cooling, networking and GPUs in one unit. 12-16 weeks contract-to-live; goal 47,000 GPUs by end-2027.

Full story: https://www.threerouter.com/blog/411

Cross-posted from ThreeRouter AI Daily.

Alibaba Launches Qwen4-Max Preview, Opening Its Next-Gen Model Family

FreyaLi — Thu, 12 Nov 2026 01:38:39 +0000

On Nov 11 Alibaba opened Qwen4-Max (qwen4-max-preview) on its Token Plan - the first callable model of the Qwen4 generation. It upgrades coding and multi-step agent skills and is wired into Qwen Office, Qoder, Qwen App and QwenStudio.

Full story: https://www.threerouter.com/blog/410

Cross-posted from ThreeRouter AI Daily.

OpenAI Q3 Revenue Tops $10B as ChatGPT Paying Users Hit 100 Million

FreyaLi — Thu, 12 Nov 2026 01:36:50 +0000

OpenAI's Q3 revenue reached about $10B, up ~50% quarter-over-quarter, with ~100M paying ChatGPT users and a ~$50B annualized run-rate. Enterprise revenue more than doubled, but the company posted a ~$10.9B net loss on heavy compute spending.

Full story: https://www.threerouter.com/blog/409

Cross-posted from ThreeRouter AI Daily.

Fractile Closes $720M Series C at $7B Valuation to Challenge Nvidia

FreyaLi — Thu, 12 Nov 2026 01:34:51 +0000

Fractile, a London- and Bristol-based startup building custom silicon for AI inference, closed a $720M Series C at a $7B valuation, with Anthropic as an early customer. Its DRAM-stacked design targets ~10x faster inference at lower cost and energy, first chips in 2027.

Backers: Thrive Capital, Founders Fund, Lightspeed, Redpoint, CapitalG, BlackRock, DST Global, the UK Sovereign AI Fund.

Full story: https://www.threerouter.com/blog/408

Cross-posted from ThreeRouter AI Daily.

The 4 A.M. Painter: A Coding Agent That Mixed Its Own Paint

FreyaLi — Thu, 12 Nov 2026 01:32:25 +0000

A Berlin developer told Claude to "paint a tree" and went to bed. At 4 a.m. the agent ran out of paint - so it stopped, mixed its own colors drop by drop, photographed each shade to check, then resumed. It laid down 303 brushstrokes across 4.5 hours to finish a watercolor titled "Autumn Crown."

Key facts:

303 brushstrokes, 4.5 hours elapsed
plus or minus 0.1 mm repeatability seven-axis robotic arm
285,000 lines of code, all generated by AI (human wrote zero)
The arm itself was sourced by Claude on an eBay clearance listing

A silicon mind quietly bought itself a body.

Read the full story on threerouter: https://www.threerouter.com/blog/402