⚡️ Kimi K3.1 Pro приблизилась к GPT-6 Astra и Claude Opus 5.5 — но стоит значительно дешевле Китайская Moonshot AI представила новую версию своей флагманской модели с контекстным окном 1 млн токенов. По результатам внутренних тестов компании, Kimi K3.1 Pro показывает впечатляющие результаты: - Terminal-Bench 4.0 — почти догнала Claude Opus 5.5. - SWE-Marathon v1.1 — обошла GPT-6 Astra и Opus 5.5. - Длительные агентские задачи — здесь преимущество пока сохраняет Claude. При этом стоимость модели составляет $2 за миллион входных и $10 за миллион выходных токенов. Ускоренная версия Highspeed обойдётся в $4 и $20 соответственно. Новинка уже доступна на kimi.ai, в Kimi Work, Kimi Code и через API. Открытые веса разработчики обещают выпустить к 4 ноября. Источник
From the original source:
Telegram — GPT Time | Нейросети | ИИ
Read full article at Telegram — GPT Time | Нейросети | ИИKimi K3.1 Pro approaches GPT-6 Astra and Claude Opus 5.5 — but costs significantly less
⚡️ Kimi K3.1 Pro approaches GPT-6 Astra and Claude Opus 5.5 — but costs significantly less. Chinese Moonshot AI unveiled a new version of its flagship model with a 1 million token context window. According to the company's internal tests, Kimi K3.1 Pro shows…
AI Summary & Key Takeaways
- ⚡️ Kimi K3.1 Pro approaches GPT-6 Astra and Claude Opus 5.5 — but costs significantly less
- Chinese Moonshot AI unveiled a new version of its flagship model with a 1 million token context window
- According to the company's internal tests, Kimi K3.1 Pro shows…
Kimi K3.1 Pro approaches GPT-6 Astra and Claude Opus 5.5 — but costs significantly less
Related stories
Block announces speech model
✔️ Block announces speech model. The company led by Twitter co‑founder and former CEO Jack Dorsey unveiled the conversational model Dessa, trained solely on human speech recordings, without text or base models. Block calls Dessa the world’s first AI conversationalist,…
Fancy harnesses make coding agents more expensive but not better, study finds
A controlled study shows that switching between coding-agent harnesses like Claude Code, OpenCode, or mini-SWE-agent barely moves pass rates more than rerunning the same setup. Cost per task, however, can triple, driven mainly by the system prompt that gets resent with every…
Russia plans to allocate 500 million rubles annually for AI and cybersecurity textbooks
📚 Russia plans to allocate 500 million rubles per year for AI and cybersecurity textbooks. The Russian Ministry of Education plans to annually direct 500 million rubles to provide students in grades 5–11 with textbooks on artificial intelligence and information security. In schools…
ElevenLabs unveils Chimera — a tool for creating realistic AI characters
🤖 ElevenLabs introduced Chimera — a tool for creating realistic AI characters. It combines the Eleven4 voice model and the Seedance 2.5 video generator. What Chimera can do: — create a character from a photo or image; — animate facial expressions; — voice text with a natural-sounding voice.