AI News: July 9, 2026

1. Cognition Releases SWE-1.7, Nearing Frontier Coding Performance at Lower Cost

Cognition. Cognition, the company behind the autonomous coding agent Devin, released SWE-1.7, which it calls its most capable model to date, trained from a Kimi K2.7 base with a refined reinforcement-learning pipeline. On Cognition’s FrontierCode 1.1 benchmark it scores 42.3%, just behind GPT-5.5 (43.0%) and Claude Opus 4.8 (46.5%) but at roughly $1.97 per task, and runs at about 1,000 tokens per second via Cerebras. The model is available immediately inside Devin. Source

2. MiniMax Plans to Open-Source a 2.7-Trillion-Parameter Model

MiniMax. The Chinese AI startup plans to open-source a 2.7-trillion-parameter model, internally called M3 Pro, potentially as early as Q3 2026. The model would be a major step up from its current flagship M3, which has 428 billion parameters, and would position MiniMax against rivals like Zhipu, DeepSeek, and Moonshot AI. Source

3. SambaNova Raises $1B at an $11B Valuation

SambaNova. The AI inference chip maker raised $1 billion in a first close of its Series F, led by General Atlantic with participation from Intel, BlackRock, T. Rowe Price, and the Qatar Investment Authority, valuing the company at $11 billion. The round follows a $350 million Series E in February 2026 and comes as JPMorganChase selected SambaNova as an on-premises inference infrastructure partner. Source

4. Prime Intellect Raises $130M to Help Enterprises Build Their Own AI Agents

Prime Intellect. The startup raised a $130 million Series A led by Radical Ventures at a $1 billion valuation, with participation from Nvidia Ventures, Intel Capital, and Dell Technologies Capital. Founded in 2024, it offers a full-stack platform of compute, reinforcement-learning frameworks, and evaluation tools that lets enterprises train their own agents, reporting a $100 million annualized revenue run rate with customers including Ramp and Zapier. Source

5. Lovable in Talks to Double Its Valuation to $13.2B

Lovable. The Swedish “vibe coding” startup is reportedly in talks to raise $300 million at a $13.2 billion valuation, double its $6.6 billion mark from December 2025, in a round expected to be led by Menlo Ventures. The platform, which lets users build software by describing it in natural language, reached a $500 million annualized revenue run rate as of June 2026. Source

6. French Startup ZML Launches a Free Cross-Chip Inference Server

ZML. The Paris-based startup released ZML/LLMD, a free LLM inference server that runs models across chips from Nvidia, AMD, Google TPU, Apple, and Intel to reduce vendor lock-in. Founded by former Zenly VP of engineering Steeve Morin, the 20-person team has raised $20 million from backers including Harry Stebbings’ 20VC and Yann LeCun. The product is free but not open source. Source

7. Kaon AI Raises $60M to Build a Personalized Story Engine

Kaon AI. The Berkeley-based startup behind the FlowGPT and Emochi generative-AI products raised $60 million from B Capital, Redpoint Ace, Goodwater Capital, and DCM. It says its consumer app Emochi has surpassed 2 million daily active users and $45 million in annual recurring revenue, with users averaging about 150 minutes per day. The funding marks a shift from its earlier FlowGPT “app store” model toward interactive AI-generated story worlds. Source

8. Bank of America Reverses Course on an OpenAI Loan Ahead of IPO

Bank of America. The second-largest US lender extended a credit line to OpenAI in recent weeks, a notable pivot toward financing money-losing AI companies. According to Bloomberg, the move is aimed at securing the bank a role in OpenAI’s anticipated public listing, with the reported roughly $520 million facility signaling that traditional banks are warming to lending against AI startups. Source