Pondero AI logo

Pondero AI

Archives
Log in
Subscribe
September 30, 2026

Pondero Brief: Sonnet 5.5 beats Opus at half price, misses more bugs

Pondero Brief - SEPTEMBER 30TH, 2026

Plus six OpenAI DevDay launches worth acting on, and the new default model in Cline.
pondero. BRIEF · SEP 30

Route coding work to Sonnet 5.5 and save Opus for the risky reviews

Anthropic's cheaper model just beat its own flagship on the coding benchmark that matters.

Anthropic's cheaper model just out-benchmarked its own flagship, 70.6% to 66.4% on Terminal-Bench 4.0, at half the price, per Anthropic. Opus 5.5 still catches more bugs on hard reviews, per CodeRabbit, so save it for those.

Also in today's brief

  • Sonnet 5.5 just out-benchmarked its own flagship
  • Cline made a new model its default overnight
 
Tools & How-To

Tool to consider · partner link

Cline switched its default model to GPT-6.1 Sol.

Cline v4.1.22 shipped September 30 with GPT-6.1 Sol as the default across a dozen-plus routed providers, including OpenAI, OpenRouter, GitHub Copilot, and Cortecs, per Cline's GitHub release notes. Claude Sonnet 5.5 and Opus 5.5 joined its recommended list in the same release. After you update, check which model your install landed on, then apply the Sonnet-first rule above.

 
Quick Hits
• Codex Security Cloud scans whole repos with no laptop open. Per OpenAI's DevDay recap, it auto-triages GitHub findings and drafts fixes now for Pro, Business, Enterprise, and Edu accounts.
• GPT-6.1 Sol shipped a week after GPT-6 Sol. OpenAI and TechCrunch report near-Astra intelligence at one-fifth the token price, the same cheap-tier bet Anthropic just made with Sonnet 5.5.
• OpenAI agents now run natively on AWS Bedrock. Per the joint OpenAI and AWS announcement, Bedrock Managed Agents ends OpenAI's Azure-exclusive run since 2019 for teams already committed to AWS.
• ChatGPT Space and Pages go after the office suite. Per OpenAI's DevDay recap, the shared workspace and human-agent documents are live now for Pro, Business, and Enterprise, with slide export coming in the following weeks.
• The Agents API can now operate software directly. Per OpenAI, computer use joined Codex's multi-agent tooling and context compaction, so teams can build computer-use agents without hosting their own execution layer.
• Dots is an always-on agent that outlives your session. Per OpenAI, it runs on GPT-6 Astra with minimal oversight for Pro and Business Premium today, so set a spend cap before anyone turns it loose.

How was today's brief?

★★★★★ Nailed it  |  ★★★ Solid  |  ★ Missed

Jonathan Hildebrandt Jonathan Hildebrandt
Co-founder and primary operator of Pondero. Writes the Pondero Brief.

Affiliate disclosure  ·  Unsubscribe  ·  Manage preferences

Pondero earns commissions on some links. This does not affect our editorial picks.

Don't miss what's next. Subscribe to Pondero AI:
← Newer Pondero Brief: Google's Gemini 4 Argon claims 77.9% on DeepSWE, pilot partners only Older → Pondero Brief: Sonnet 5.5 lands 2 points off Opus on GDPval, at half the price
pondero.ai
Bluesky
LinkedIn
Twitter
LinkedIn
Powered by Buttondown, the easiest way to start and grow your newsletter.