DeepSeek's Budget Model Just Beat Its Own Flagship on Agent Benchmarks
DeepSeek retrained its 284B flash model and it now outscores the company's own flagship V4-Pro-Preview across all nine agent benchmarks — including a 645% jump on software engineering tasks — at a fraction of the cost. »
Alibaba's Qwen 3.7 Flash: 1M-Context Vision AI at $0.03 per Million Tokens
Alibaba quietly released Qwen 3.7 Flash on July 27 — a vision-language model with a 1M-token context window priced at $0.03 per million input tokens, making it the cheapest multimodal high-context model on the market. »
MCP Goes Stateless: The Biggest Model Context Protocol Revision Since Launch
The MCP 2026-07-28 release candidate removes sessions, drops the initialization handshake, and hardens OAuth 2.1 authorization — calling it 'the largest revision since launch.' Every AI agent stack that runs MCP in production needs to take note. »
Valve's Steam Frame Clears FCC — Everything Is Now in Place for Launch
Valve's standalone VR headset received FCC equipment authorization on July 29 — the last regulatory hurdle before a retail announcement. With stock already in the US and a live game storefront, launch is no longer a question of if. »
Immersed Visor Raises Its Price $200 Tomorrow — and Finally Looks Close to Shipping
The work-focused VR headset jumps from $500 to $700 for new pre-orders on July 31, driven by Qualcomm's chip price hike. More interestingly: Immersed says it's locking in production timelines and will share a shipping update in August. »