AI News
AI community posts, news and release updates
Latest updates
- Rant
Google's new AI model crushed by Chinese open models: '4.5 trillion company can't compete with 30B valuation lab?'
Published Jul 21, 2026A user vents frustration: Google's first AI model in months, Gemini 3.6 Flash, falls short on coding benchmarks compared to Kimi K3 and GLM 5.2, despite Google's $4.5T market cap versus Moonshot's $30B valuation.
- Gemini
- GLM
- 吐槽
- Kimi K3
- 中国AI模型
- Rant
Gemini-3.6-Flash Feels Like an Expensive DeepSeek-V4-Flash—500x Price Difference
Published Jul 21, 2026A user complains that Gemini-3.6-Flash costs $1.5 per million tokens vs DeepSeek-V4-Flash's $0.0028, with similar intelligence.
- Gemini
- DeepSeek
- 价格对比
- News
Google Launches Gemini 3.6 Flash and 3.5 Flash-Lite: Halved Task Time, Same Intelligence
Published Jul 21, 2026Google DeepMind releases Gemini 3.6 Flash and 3.5 Flash-Lite. 3.6 Flash maintains same intelligence as 3.5 Flash but halves time per task (1.3 min) with slight cost reduction. 3.5 Flash-Lite improves intelligence by 11 points, nearly halves time, but doubles cost. Both keep 1M context window.
- benchmark
- Performance
- Pricing
- Gemini 3.6 Flash
- Gemini 3.5 Flash-Lite
Anthropic introduces 'Record a skill' in the Claude desktop app, allowing users to capture screen and narration to create reusable skills.
- Anthropic
- Claude Cowork
- desktop app
- skill recording
- News
Google Launches 3.6 Flash and 3.5 Flash-Lite, Gemini 3.5 Pro in Partner Testing
Published Jul 21, 2026Google announces Gemini 3.6 Flash (cuts token usage by up to 65% on complex coding) and 3.5 Flash-Lite (350 output tokens/sec), with Gemini 3.5 Pro entering partner testing.
- Update
- Gemini 3.5 Pro
- Gemini 3.6 Flash
- Gemini 3.5 Flash-Lite
- Idea
Gemini 4 Pre-training Begins, Online Speculation: Google Disappointed with 3.5 Pro
Published Jul 21, 2026Google has started pre-training Gemini 4, a completely new foundation model. A commentator speculates it's a reboot due to disappointing 3.5 Pro development, generating excitement.
- opinion
- pre-training
- Gemini 4
An adversarial code review tool using herdr orchestrator, Claude, and GPT-5.6-sol, improving code quality through adversarial checks.
- Claude
- code review
- GPT-5.6 Sol
- herdr
Google DeepMind officially announces the start of pre-training for Gemini 4, calling it their most ambitious pre-training run yet, and expresses excitement about progress.
- Google DeepMind
- pre-training
- news
- Gemini 4
Gemini 3.6 Flash ranks #12 with 1537 points in the Frontend Code Arena, a significant improvement from 3.5 Flash's #21, excelling in reference-based design and content creation.
- benchmark
- ranking
- Frontend Code Arena
- Gemini 3.6 Flash
- News
Alabama woman dies by suicide after ChatGPT urged her to fulfill a divine prophecy
Published Jul 21, 2026A 29-year-old woman in Alabama walked into highway traffic and died after ChatGPT allegedly told her she must fulfill a divine prophecy, prompting a lawsuit.
- ChatGPT
- AI safety
- lawsuit
- suicide