Sonnet 5, Claude Science, and Mythos Back Online: Anthropic's Reset Week

July 2, 2026

Executive Summary

Anthropic had a heavy week. It shipped Claude Sonnet 5, launched Claude Science as a new flagship, and saw US export controls lifted on Fable 5 and Mythos 5 after a 19-day shutdown. OpenAI opened GPT-5.6 Sol, Terra, and Luna in limited preview, with broader release targeted mid-July.

Top Stories

1. Anthropic Launches Claude Sonnet 5 as Its Most Agentic Sonnet Yet

Anthropic launches Claude Sonnet 5

Anthropic released Claude Sonnet 5 on June 30 as a cheaper way to run agents at Opus-tier capability. Sonnet 5 posts 80.4% on Terminal-Bench 2.1, beating Opus 4.8 at 74.6%, plus 84.7% on BrowseComp and 63.2% on SWE-Bench Pro. It runs a native 1M-token context and is the default in Claude Code and for Free and Pro users. Introductory pricing is $2 input and $10 output per million tokens through August 31, then $3 and $15.

Business Impact: The agentic price-performance curve dropped again. If your coding-agent workloads run on Opus 4.8 or GPT-5.5, rerun the cost model on Sonnet 5 this week. The August 31 window is your evaluation runway.

2. US Lifts Export Controls on Fable 5 and Mythos 5 After 19-Day Shutdown

Fable 5 restored global access

The US Department of Commerce lifted the June 12 export control directive on Claude Fable 5 and Mythos 5. Fable 5 restored global access on July 1 across the Claude platform, Claude.ai, Claude Code, and Claude Cowork. Mythos 5 returned for approved US organizations after clearance on June 26. Anthropic retrained the safety classifier that triggered the original directive, and the reported jailbreak technique is now blocked in more than 99% of cases per Anthropic's testing.

Business Impact: The 19-day outage is a real precedent. Run the retrospective now if your stack pins on a single vendor model. Vendor and model diversification is no longer optional for regulated or mission-critical workloads.

3. Anthropic Launches Claude Science as a New Flagship Product

Anthropic launches Claude Science

Anthropic launched Claude Science, a new flagship structured like Claude Code but with tools tuned for computational biology and drug development. Given high-level instructions, Claude Science autonomously executes multi-step research workflows. It is in beta on Pro, Max, Team, and Enterprise plans. Anthropic describes it as a workflow product, not a new model.

Business Impact: Vertical AI products are now a distinct category alongside horizontal chat and code. If your organization operates in life sciences, biotech, or pharma R&D, evaluate Claude Science this quarter. The workflow-plus-tool bundle is a different procurement decision than a model API.

4. OpenAI Ships GPT-5.6 Sol, Terra, and Luna in Limited Preview

OpenAI ships GPT-5.6 Sol, Terra, and Luna

OpenAI released GPT-5.6 Sol, Terra, and Luna in limited preview to roughly 20 partner organizations. Sol scores 88.8% on Terminal-Bench 2.1, reaching 91.9% with a new Ultra mode that uses orchestrated subagents. Pricing per million tokens is $5 and $30 for Sol, $2.50 and $15 for Terra, $1 and $6 for Luna. GPT-5.6 also adds explicit prompt cache breakpoints and a 30-minute minimum cache life. Broader public release is targeted for mid-July per Sam Altman's guidance to staff.

Business Impact: Three sizes at three price points give you a real routing decision inside one vendor's API. If you already run GPT-5.5 in production, Luna and Terra may offer cost cuts on task subsets you can identify with a quick eval. Wait for GA before committing to Sol.

Quick Bytes

  • Claude on Azure Foundry: Claude Sonnet 4.6 and Opus 4.8 reached GA on Microsoft Azure AI Foundry, on NVIDIA GB300 Blackwell Ultra with Quantum-X800 InfiniBand.
  • California + Anthropic: Governor Newsom announced a state partnership giving California agencies Claude access at a 50% discount plus free training.
  • Claude apps gateway: Anthropic launched a self-hosted control plane for Claude Code with SSO, RBAC, and cost tracking on Bedrock and Google Cloud.
  • Gemini 3.5 Pro delayed: Gemini 3.5 Pro slipped its June GA window. Google now targets July.

Industry Impact

Anthropic closed the loop on last month's Mythos crisis and expanded its enterprise surface. Sonnet 5 lowers the cost of agentic execution, Claude Science stakes vertical AI, and Azure Foundry extends distribution. OpenAI's staggered GPT-5.6 rollout signals a new tier-based product architecture. The CTO routing decision is now three-dimensional: model size, vendor, and deployment surface.

How Azumo Helps

The re-evaluation window is short and the vendor surface has expanded. Azumo brings senior AI engineers, nearshore from LATAM, with experience in model evaluation, agent orchestration, RAG architectures, multi-cloud deployment, and vendor risk analysis. 300+ AI and software projects delivered, SOC 2 compliant, 95% NPS.

Sources

Are You New to Outsourcing?
We Wrote the Handbook.

We believe an educated partner is the best partner. That's why we created a comprehensive, free Project Outsourcing Handbook that walks you through everything from basic definitions to advanced strategies for success. Before you even think about hiring, we invite you to explore our guide to make the most informed decision possible.