Loading / 加载中
Thinkgap feed
During his final earnings call as Apple CEO, Tim Cook emphasized that the company's hybrid AI strategy, combining on-device and cloud processing, is a competitive advantage. He contrasted this with rivals that depend heavily on massive cloud infrastructure. Cook steps down on September 1.
OpenAI has reduced prices for its GPT-5.6 model family. The Luna variant becomes 80% cheaper, while the Terra variant sees a 20% price cut. The Sol Fast inference service is now up to 2.5× faster. These updates are seen as a direct competitive move against Chinese AI models. Meanwhile, Grok 4.5 High remains on the Pareto frontier of cost-performance.
A user of the 'opencode go' subscription service received an HTTP 403 error when attempting to use DeepSeek V4 Flash. The error body is a RegionError stating that the latest version of the model is only available hosted in China and requires explicit opt-in. This indicates that DeepSeek has geo-restricted the V4 Flash model to Chinese IP addresses or accounts. The restriction affects non-China users even when using a third-party service that previously provided access. The situation may signal a change in DeepSeek's deployment policy for its latest models.
OpenCode, an open-source coding agent launched in June 2025, has grown to 13 million monthly active users and an annualized revenue approaching $60 million, with 160,000 paid subscribers and an inference business processing 7 trillion tokens daily. Its growth was accelerated in January 2026 when Anthropic blocked Claude Code usage via OpenCode, which inadvertently raised awareness and led OpenAI to grant official support. The company occupies a neutral, model-agnostic position, supporting over 70 models and benefiting from fierce model competition. Its inference unit achieves profit margins up to 80-90%, and it uses free tokens as customer acquisition cost. OpenCode’s co-founders emphasize product judgment over speed, warn against feature bloat, and have rejected acquisition offers to pursue long-term independence.
The release of Chinese open-weight models Kimi K3 and Qwen3.8 Max has sparked fierce debate on AI commoditization, cost structures, and U.S. policy. The article argues that intelligence is becoming a commodity where competitive advantage hinges on unit cost of inference, not just token price. Chinese labs benefit from distilling frontier models, narrowing the gap faster, while their open-weight strategy aims to commoditize AI to boost China's manufacturing and robotics edge. The analysis warns that U.S. restrictions on using frontier models for cybersecurity force defenders to rely on Chinese models, and it proposes legal reforms to permit distillation and clarify fair use for model training.
ThinkyMachines has announced Inkling-Small, an open omni-modal model capable of processing audio, text, and image inputs. The model is described as a frontier open Omni. It is available to run on Inference Endpoints, simplifying deployment for developers. This release expands open access to multimodal AI models beyond text and image to include audio.