The Intake
The Intake — Monday, August 18, 2026
On the substrate
Stripe acquires OpenRouter for more than $7 billion
Bloomberg, Aug 16 TechCrunch, Aug 16 Dataconomy, Aug 17
If you route model API calls through OpenRouter, the service's ownership changed on August 16. Stripe reportedly agreed to acquire OpenRouter for more than $7 billion. OpenRouter routes requests to more than 400 models through a single API endpoint. Providers include Anthropic, OpenAI, Google, Meta, and DeepSeek.
The service has roughly 8 million registered developers. OpenRouter's Series B valuation was $1.3 billion, set in May 2026. The acquisition price is approximately 5.4x that figure. CEO Alex Atallah has described OpenRouter as doing for AI what Stripe does for payments. Stripe declined to comment to TechCrunch.
No product or integration changes have been announced. If OpenRouter is in your routing stack, ownership changed; no service disruptions have been reported.
Google releases Gemini 3.7 Flash
Google DeepMind model card APIdog, Aug 13 BenchLM.ai
Gemini 3.7 Flash is Google's Flash-tier model for coding and agent automation workloads. Google released version 3.7 on August 13, 23 days after Gemini 3.6 Flash.
On DeepSWE v1.1, 3.7 Flash scores 65.3%. The prior model (3.6) scored 49.0%. FrontierCode 1.1 is now 43.6%; the prior model scored 34.4%. AutomationBench is now 30.4%, up from 17.0%. WebDev Arena Elo is now 1588, up from 1538. The context window is 1,048,576 tokens. Artificial Analysis reports approximately 340 output tokens per second.
Introductory pricing is $0.75 per million input tokens and $3.75 per million output tokens. Both rates apply through December 31, 2026, and double on January 1, 2027. If you're optimizing cost alongside capability, the January pricing change is the constraint to plan around.
OpenAI previews GPT-5.6 Sol Ultrafast at up to 750 tokens per second
OpenAI Cerebras investor release, Aug 13 HPCwire
If inference latency is the bottleneck in your agent pipelines and you're running GPT-5.6 Sol, OpenAI announced a limited preview of Ultrafast mode on August 13. OpenAI says the mode delivers up to 750 output tokens per second — approximately 14x standard throughput.
The mode runs on Cerebras hardware. The model weights and context window are the same as GPT-5.6 Sol Standard. The preview is waitlist-only. Pricing and a general availability date are not yet confirmed.
OpenAI's initial test cohorts include coding and financial research applications. Voice AI and e-commerce companies are also in the cohort. Ultrafast mode runs on Cerebras hardware — if your deployment has constraints around inference provider selection, that's the tradeoff this mode introduces.
---
For operators
Amazon closes Bedrock Agents Classic to new accounts; AgentCore is the migration path
AWS documentation Enterprise DNA ServerGurus
If you're building new agents on Amazon Bedrock Agents, the service entered maintenance mode on July 30. Amazon renamed it Bedrock Agents Classic and closed it to new accounts on that date. CreateAgent and InvokeInlineAgent calls from accounts without prior usage now return AccessDeniedException (HTTP 403).
Existing agents continue running on allowlisted accounts. The model catalog in Classic is frozen at July 30. Models released after that date are available only through AgentCore.
AWS recommends migrating to Amazon Bedrock AgentCore. AgentCore supports the Claude Agent SDK and OpenAI Agents SDK. Strands, LangChain, and custom frameworks are also compatible. No migration deadline has been announced. An automated migration guide is available in the AWS agent toolkit at github.com/aws/agent-toolkit-for-aws. If you're using Bedrock Agents for new project work, AgentCore is the current path. The migration guide handles the transition automatically.
---