Mira Murati's lab, Thinking Machines, released Inkling-Small on July 15, a Mixture-of-Experts model with 276 billion total parameters and 12 billion active at inference, making it the first US-based lab to ship a frontier-class open-weights model competing with Chinese releases on capability and price. It performs at or near the level of models four times its size on agentic benchmarks, scoring 80.2 percent on SWE-Bench Verified, 64.7 percent on Terminal Bench 2.1, and 95.1 percent on AIME 2026 at maximum effort. Serverless inference on the Tinker API runs $0.30 per million input tokens and $1.20 per million output tokens at 256K context, roughly half the cost of OpenAI’s Luna and well below Kimi K3’s $3.00/$15.00, while DeepSeek’s V4-Flash-0731 remains cheaper at $0.14/$0.28. Inkling-Small offers a 1-million-token context window and native multimodal support. The two-model strategy mirrors DeepSeek's Flash and Pro split, with the larger Inkling (975B total, 41B active) handling high-complexity reasoning. Fine-tuning on Tinker costs $1.73 per million tokens for 64K context, currently at a 50 percent introductory discount.