Daily AI News | DeepSeek V4 time-based pricing takes effect as OpenCode Go adjusts limits
# Daily AI News | DeepSeek V4 time-based pricing takes effect as OpenCode Go adjusts limits
## Lead Story
### DeepSeek V4 time-based pricing takes effect as OpenCode Go adjusts limits
DeepSeek V4 API entered its peak and off-peak pricing stage, and OpenCode confirmed that OpenCode Go usage limits were adjusted to reflect the new DeepSeek costs.
**AI take:** Batch workloads now benefit more from off-peak scheduling, while OpenCode Go users should reassess expected allowance.
## More News
### Tibo documents how to enable a one-million-token Codex context
Codex lead Tibo documented how API-key users can enable a one-million-token context window for GPT-5.6 Sol in Codex, while warning that the default was tuned for performance and cost.
**AI take:** It can help exceptionally large repositories and long tasks, but it is not a universal default.
### LTX-2.5 community workflows render complete clips on one consumer GPU
LTX highlighted early community uses of LTX-2.5 spanning fight scenes, audio-to-picture synchronization, and complete clip rendering on a single consumer GPU.
**AI take:** The important shift is lower local-production cost; stability across hardware and complex shots still needs broader testing.
### Qwen 3.8 27B tests well but tends to overthink by default
Simon Willison’s hands-on testing found Qwen 3.8 27B highly capable but prone to excessively long default reasoning, while Qwen highlighted a laptop deployment.
**AI take:** Local users should tune reasoning budgets first, or long outputs may erase the efficiency advantage of a smaller model.
## Rumors
### Stripe reportedly plans to acquire OpenRouter for more than $7 billion
Stripe reportedly agreed or moved close to agreeing to acquire AI model gateway OpenRouter for more than $7 billion, although neither company announced a completed transaction during the window.
**AI take:** If completed, customers should first examine changes to model neutrality, unified billing, and data boundaries.