signal: week of 2026-07-10
2026-07-10
https://tinksoft.com/log/signal-2026-07-10/
--------------------------

Three major model families shipped within 48 hours of each other this week, and Google decided to scrap theirs and start over.

- [OpenAI launches GPT-5.6 with Sol, Terra, and Luna tiers](https://openai.com/index/previewing-gpt-5-6-sol/) — OpenAI's new three-tier model family went to general availability on July 9: Sol ($5/$30 per 1M tokens) for frontier reasoning and agentic work, Terra at roughly GPT-5.5 performance for half the cost, and Luna as the budget speed option.

- [Anthropic redeploys Claude Fable 5 after export-control suspension](https://www.anthropic.com/news/redeploying-fable-5) — Fable 5 came back online July 1 after a 19-day suspension triggered by US export controls; Anthropic added a new cybersecurity classifier that blocks the reported jailbreak technique in over 99% of cases.

- [SpaceXAI launches Grok 4.5 as a coding-first model](https://x.ai/news/grok-4-5) — xAI's first release since acquiring Cursor: a 1.5T-parameter model priced at $2/$6 per 1M tokens with a 500k-token context window, pitched as "Opus-class but faster and cheaper."

- [Meta ships Muse Spark 1.1 and opens the Meta Model API](https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/) — Meta's first paid API model ($1.25/$4.25 per 1M tokens) is a multimodal reasoning model built for agentic tasks, with a 1M-token context window and major gains in tool use and coding.

- [Google delays Gemini 3.5 Pro to July 17 for a full architectural rebuild](https://hackernoon.com/google-delays-gemini-35-pro-to-july-17-the-strategic-play-behind-the-scrapped-base-model) — DeepMind scrapped the 2.5 Pro base model entirely after enterprise testers flagged coding performance gaps and long-task reasoning issues, opting for a ground-up redesign targeting math reasoning and a 2M-token context window.

- [Anthropic launches Claude Science for drug discovery](https://www.cnbc.com/2026/06/30/anthropic-launches-ai-drug-discovery-program-claude-science.html) — A dedicated AI workbench integrating 60+ scientific databases and tools, with an internal drug discovery program targeting neglected diseases and up to $30k in research grants per project.

- [Anthropic releases Claude Sonnet 5 at introductory pricing](https://www.anthropic.com/news/redeploying-fable-5) — Sonnet 5 launched June 30 with near-Opus 4.8 performance at $2/$10 per 1M tokens (introductory rate through August 31), becoming the default model across all Claude plans.