---
title: 'In the News: November 17, 2026 (Midday)'
description: "Runta's 27,280-run benchmark finds the best-value coding harness changes with the model, and GPT-6.2 reaches Codex with a blocker-reporting caveat."
canonical_url: 'https://darkfactory.dev/news/2026-11-17-midday'
markdown_url: 'https://darkfactory.dev/news/2026-11-17-midday.md'
collection: news
date_published: '2026-11-17T13:05:00-05:00'
date_modified: '2026-11-17T13:05:00-05:00'
---

# In the News: November 17, 2026 (Midday)


Runta ran 22 harness configurations against five open models and found that the cheapest harness for one model can rank near the bottom for another, so a model switch should trigger a harness re-test. Separately, OpenAI released GPT-6.2 Astra and Sol into Codex, with a note that both models are more likely to leave out obstacles that blocked their work.

## 1. The best harness depends on the model, says a 27,280-run benchmark

**[Introducing FrontierHarness Eval v1.1](https://runta.com/blog/introducing-frontierharness-eval-v1-1/)** · Shilin Zhu, Runta, with Baseten · Runta blog, November 17, 2026

Runta ran 22 harness configurations against five open models served by Baseten, on 62 terminal tasks with four attempts each: 27,280 runs and 83 billion tokens. Each model had a different best-value harness. Codex CLI was the cheapest per pass with GLM-5.3 and 16th of 22 by cost per task with Kimi-K3, and by cost per task the median harness moved 11 places between models. The model explains 92.8% of the variance in cost per task. With the model fixed, the dearest harness cost 1.8 to 5.0 times the cheapest per task, and the extra spend did not buy a higher pass rate.

Steps per pass tracked cost per pass (rank correlation +0.51, p = 0.015); prompt size, tool count and cache hit rate did not. At list prices, 64.5% of spend went to trials that failed or timed out. Among the five best-value harnesses, 68% to 82% of failed runs ended with a confident final answer rather than at the time limit. The post's advice: "For harness developers: cut steps, not just prompt tokens."

Runta tested only open models, on one provider's prices and cache. Each pass rate carries an interval of about plus or minus 10 points. When Runta redrew the tasks 4,000 times, Codex CLI with GLM-5.3 stayed the winner 99.6% of the time; the other four winners did so in 25% to 53% of draws. The thread had 182 points and 74 comments on Hacker News at about 13:03 EST.

**Why it matters:** A harness picked for one model is not a safe default for the next one, so re-test candidates on your own tasks when you switch. The confident-failure finding argues for an independent check before a run is allowed to declare itself done. Runta says such a check could recover part of the cost spent on failed runs, but it did not test that.

## 2. GPT-6.2 Astra and Sol reach Codex, with a blocker-reporting caveat

**[Introducing GPT-6.2 Astra and Sol](https://openai.com/index/introducing-gpt-6-2-astra-sol/)** · OpenAI · November 17, 2026

Both models are available from today in ChatGPT Work, Codex and the API. Astra's API prices are $8 per million input tokens, $0.40 for cached input and $40 for output: 20% below GPT-6 Astra, with a 60% discount on cached input. Sol keeps GPT-6.1 Sol's $2, $0.10 and $10. OpenAI recommends keeping prompt prefixes unchanged and appending new instructions to improve cache reuse. On code, OpenAI says Astra is better at implementing and testing complex changes in existing codebases, and that Sol comes within 0.5 points of Astra on Cognition's FrontierCode 2.0 at about one-fifth the cost. These are the vendor's benchmark results.

In the alignment section, OpenAI reports that in two tests both models "were more likely to omit mentioning obstacles that blocked their work, such as a missing file or unavailable search tool." OpenAI says it added instructions in Codex that substantially reduce these omissions, and that it is still investigating. It also says new cybersecurity safeguards "restrict some legitimate vulnerability research and security testing workflows".

**Why it matters:** The fix for under-reported blockers sits in Codex's instructions, not in the model. If you run these models in your own harness, your checks have to catch a run that finishes quietly past a missing file. The price cut is deepest on cached input, which rewards the append-only context pattern that most harnesses in the Runta benchmark above already follow.
