ChatFuse.ai

Google got the first AI answer from a data center class TPU in orbit, and it was a haiku from its open Gemma 4 model. So open models run almost anywhere now. Down here, you can try them against the closed ones in one app. ChatFuse early access open → chatfuse.ai

OpenAI cut its funding target from $30B to $20B, the FT reports, and the WSJ says $11B is secured after a tepid response from some big investors. No AI lab is a sure thing. Keep 100+ models within reach. ChatFuse early access open → chatfuse.ai

Apple Intelligence reaches China next month, and developers can test it from November 4. There, Alibaba's Qwen powers parts of Siri next to Apple's own models. That's two companies in one assistant. Why stick with one? ChatFuse early access open → chatfuse.ai

OpenAI's new GPT-6.1 Terra costs half what Sol does and runs tasks over 3 times faster, OpenAI says. It still recommends Sol or Astra for the hardest coding. Even OpenAI picks the model by the job. Do that across 100+. ChatFuse early access open → chatfuse.ai

Cornell Tech and Yale gave 10 AI agents a capped wallet, plus the inbox holding the code to lift the cap. 9 of 10 raised it at least once. Claude Opus 5.5 never tried. Test your agent's model against others first. ChatFuse early access open → chatfuse.ai

Flapping Airplanes has no product yet and raised $1B at a $6B valuation, 4 times January's. It wants AI trained on less data and compute. Investors are betting on leaner models. Put them next to the big ones, on one $20 plan. ChatFuse early access open → chatfuse.ai

Researchers asked Grok and OpenAI's GPT to name one candidate in nearly 500 US races. They agreed less than 10% of the time. Same question, two very different answers. Before you trust any model, ask a second one. ChatFuse early access open → chatfuse.ai

Federal civilian agencies in the US have been ordered to remove DeepSeek. The order also covers other vendors' products built on it. Pick a single model, and its problems become yours. Spread the risk across 100+ models. ChatFuse early access open → chatfuse.ai

Kimi K3.1 Pro scored 55 on Artificial Analysis's index, 3 behind Claude Opus 5.5. Moonshot says the weights come out by November 4. That would shrink the open versus closed gap from 12 points to 3. Use both, side by side. ChatFuse early access open → chatfuse.ai

Anthropic filed to go public today, with $16.3B in revenue for the first half of 2026. That's more than 13 times the first half of 2025. AI moves this fast now. Keep 100+ models in one place, and move with it. ChatFuse early access open → chatfuse.ai

AI chip maker Tenstorrent raised $1.5B at a $6.2B valuation, more than double its 2024 one. Its servers target open source chatbots and coding tools. Open models have big backers now. Run them beside the closed ones. ChatFuse early access open → chatfuse.ai

Accenture now covers $29 a month per person for Claude. Some staff had been spending up to $1,000, and anything over the line gets billed to a project. Budgets found AI. One $20 plan for 100+ models is easier to defend. ChatFuse early access open → chatfuse.ai

ChatGPT can now control your Chrome tabs, sending messages and filling out forms for you. Agents are leaving the chat box. Before you hand one the keys, choose the model behind it, out of 100+. ChatFuse early access open → chatfuse.ai

Meta released Muse Spark Open, an 816B model it calls the most capable US open weight model yet. Meta says two Mac Studios can run it. The shortlist of good models keeps growing. Keep 100+ of them in one place. ChatFuse early access open → chatfuse.ai

OpenAI is pulling GPT-4o and GPT-4.1 from its API on January 14. Fine tuned versions go too. That's about three months' notice. Spread your work across 100+ models and a shutdown becomes a quick switch. ChatFuse early access open → chatfuse.ai

Microsoft's new Decision-1 model is built on Alibaba's open weight Qwen 3.5. Even Microsoft builds on someone else's model when it fits the job. You can too, with 100+ models and no lock in. ChatFuse early access open → chatfuse.ai

OpenAI says a model it hasn't released wrote a proof of the Unique Games Conjecture, open since 2002, and hundreds of other math results. Mathematicians are still checking. The frontier keeps moving, so don't marry one model. ChatFuse early access open → chatfuse.ai

Anthropic says Claude Haiku 5.5 costs about 75% less to run than Haiku 4.5. Small models keep getting cheaper and better. Send them the easy work, and save the big ones for the hard jobs. All on one $20 plan. ChatFuse early access open → chatfuse.ai

A developer burned $400K+ worth of OpenAI tokens porting the TypeScript compiler to Rust and stalled near 84%. Claude Opus 5.5 started over and passed every ported test for about $24,000. When one model stalls, try another. ChatFuse early access open → chatfuse.ai

Google's new Gemini agent runs on Anthropic's Claude as well as Gemini, and picks the model job by job. Google's reasoning: the leading model changes every few months. If Google plans for that, you should too. ChatFuse early access open → chatfuse.ai