◆ NOISE IN → SIGNAL OUT◆ READALCHEMIST.COM◆ FREE / NO PAYWALL◆ NOISE IN → SIGNAL OUT◆ READALCHEMIST.COM◆ FREE / NO PAYWALL
THE DIGITAL ALCHEMIST
AIIMPACT 80

OpenAI's GPT-6.2 Launch Skips the Benchmark Where It Loses

OpenAI's charts show GPT-6.2 Astra beating its own predecessor nine times out of nine. The independent scoreboards tell you it still trails Anthropic, and power users are calling it confusing.

2026-11-186 MIN READ#OpenAI · #GPT-6.2 · #Anthropic · #AI benchmarks · #frontier models
The Digital Alchemist
The Digital Alchemist

OpenAI's launch post for GPT-6.2 Astra and Sol reads like a highlight reel because it is one. Every chart shows the new models winning. What the page does not show is the one scoreboard OpenAI does not control.

This is not a capability story. It is a trust story, and OpenAI keeps failing the trust part.

On Tuesday, Nov. 17, OpenAI shipped GPT-6.2 Astra at $8 per million input tokens and $40 per million output tokens—20% below its predecessor—and GPT-6.2 Sol at $2 and $10. In OpenAI's own charts, Astra beats GPT-6 Astra on all nine listed benchmarks, and Sol beats GPT-6.1 Sol on eight of nine. OpenAI grades itself, and it passes.

Artificial Analysis, the independent benchmark outfit OpenAI does not own, landed somewhere less flattering. Astra scored 62 on its Intelligence Index, five points above GPT-6 Astra but five points below Anthropic's Claude Fable 5.2, which ranked first. Sol scored 57—matching what GPT-6 Astra scored months ago, at a fraction of the cost.

It is a real improvement. It is also a model OpenAI calls its new frontier flagship, outscored by a rival on the index built to compare across labs.

On Terminal-Bench 4, the broader agentic-coding suite, Astra placed 5th and Sol 7th. On Terminal-Bench-Science, the narrower benchmark OpenAI chose to headline, Astra leads outright. When a company picks the test it wins and skips the leaderboard it loses, that is marketing wearing a lab coat.

The Digital Alchemist
The Digital Alchemist

VentureBeat ran its launch coverage under the headline "Essentially unusable," reporting negative reception from AI power users. Simon Willison, the industry's closest thing to a neutral referee, called the early buzz mixed and noted that Astra 6.2 isn't scoring as high as Fable 5.2 on various benchmarks, adding complaints about confusing communication. On Hacker News, one widely upvoted comment: benchmarks can be trusted or not, but Astra is roughly 20% slower and less powerful than Fable while costing more per task than Sol.

The cadence is the tell

Four major OpenAI model drops since Sept. 3: GPT-6 Astra that day, GPT-6 Sol and Luna on Sept. 22, GPT-6.1 Sol on Sept. 29, and GPT-6.2 Astra and Sol on Nov. 17. Four launches in eleven weeks is not the rhythm of a lab coasting on an uncontested lead. It is the rhythm of a lab that keeps getting caught by the market before its own PR cycle finishes.

Translation: we found more efficiency, not more intelligence, and we need you to not notice the difference.

If you pick Astra over Sol because OpenAI's chart says so, you are trusting a chart OpenAI built, scored, and published without outside verification. Independent numbers say Astra is a few points ahead of Sol on hard reasoning and several points behind Claude on the index that compares labs. Run your own eval on your own task before you pay the Astra markup.

What to watch: whether OpenAI publishes a benchmark it lets a third party run cold, whether the next Anthropic or Google release ships with reproducible third-party numbers attached from day one, and whether Artificial Analysis's gap between Astra and Fable 5.2 widens or closes as more labs test.

Artificial Analysis Intelligence Index, independent score
Claude Fable 5.265ptsGPT-6.2 Astra62ptsGemini 4 Carbon60ptsClaude Opus 5.559ptsGPT-6.2 Sol57pts
Source: Artificial Analysis, Nov. 17, 2026.
Sources
  1. Introducing GPT-6.2 Astra and Sol
  2. GPT‑6.2 Astra and GPT‑6.2 Sol - Simon Willison's Weblog
  3. 'Essentially unusable': OpenAI launches GPT-6.2 Astra and Sol to negative reception from AI power users
  4. OpenAI has the #2 model in the Intelligence Index with the release of GPT-6.2 Astra and GPT-6.2 Sol
  5. OpenAI releases GPT-6.2 Astra and Sol with lower API prices
  6. GPT-6.2 Is Here: Astra and Sol, Who Gets Them, and What They Cost
  7. OpenAI launches GPT-6.2 Astra and Sol with lower API prices and some benchmark gains
  8. GPT-6.2 Astra and Sol Launch: Benchmarks and Pricing
  9. OpenAI Launches GPT-6.2 Astra and Sol in ChatGPT Work, Claiming It Beats Claude Fable 5.2 at 5% of the Cost
  10. OpenAI Surprise-Launches GPT-6.2 Astra & Sol on Anthropic’s IPO Date: Competitive Strike With Task Cost Only 5% of Claude to Reshape the Generative AI Market
  11. OpenAI launches GPT-6.2 Astra and Sol with major performance gains at a fraction of the cost - Neowin
  12. Simon Willison’s Weblog
  13. GPT‑6.2 Astra and Sol
  14. GPT-6.2 Astra and Sol: full price comparison
  15. GPT‑6 Astra
  16. GPT-6.2 Astra and Sol: what's new, prices and how to use them
  17. OpenAI on X: "GPT-6.2 Astra—a frontier model built for your hardest projects. It sets new highs on evals across professional work and scientific research at less than a third of the cost per task of GPT‑6 Astra. You’ll see the difference in things like slideshows, websites, and 3D designs." / X
  18. GPT-6.2 Astra and GPT-6.2 Sol—Models built for the real world - YouTube
  19. OpenAI on X: "Some of you have been asking very politely. Some of you have just been asking. GPT-6.2 Astra and Sol are now available. Super efficient and built to perform where it counts—the real world. https://t.co/qr2iP0GpfX" / X
  20. Claude Opus 5.5, GPT-6 Sol, GPT-6 Luna, and a new price war
  21. GPT-6.1 Sol - Learn AI - Miraheze
  22. OpenAI Unveils GPT-6.1 Sol at DevDay With New Codex and ChatGPT Tools
  23. GPT-6.1 Sol: Release Date, Availability and Changes
  24. OpenAI releases GPT-6.1 Sol at DevDay 2026
  25. GPT-6.1 Sol: Features, Benchmarks, Pricing, and Access
  26. ChatGPT 6.1 Sol: release date, price and when to use it
  27. GPT-6.1 Sol Guide (Oct 2026): Ultrafast Pricing, Codex, API
  28. GPT-6.1 Sol: Benchmarks, Pricing and Specs (2026)
  29. OpenAI & ChatGPT Timeline: GPT Release Dates to GPT-6.1 (2026)
  30. GPT-6.1 Sol — Benchmarks, Specs & Release Date
← back to the feed
QueryFlow
ASK

Ask your database anything.

Stop writing queries. You don't need to know SQL to get the answer. QueryFlow AI writes it for you.

Native Mac app · by yForest

  • Snowflake
  • Redshift
  • PostgreSQL
  • MySQL
  • BigQuery
  • Databricks
  • Salesforce
  • Google Sheets
  • CSV/Excel
NVDA 150.96 ▼ 2.89%MSFT 508.29 ▲ 0.40%GOOGL 282.44 ▲ 0.80%AMZN 267.74 ▲ 3.82%META 616.38 ▼ 0.40%TSLA 383.72 ▲ 0.95%AMD 530.14 ▲ 5.31%AVGO 354.05 ▲ 3.64%PLTR 156.67 ▲ 0.07%COIN 110.51 ▼ 3.92%MSTR 90.32 ▼ 6.40%NVDA 150.96 ▼ 2.89%MSFT 508.29 ▲ 0.40%GOOGL 282.44 ▲ 0.80%AMZN 267.74 ▲ 3.82%META 616.38 ▼ 0.40%TSLA 383.72 ▲ 0.95%AMD 530.14 ▲ 5.31%AVGO 354.05 ▲ 3.64%PLTR 156.67 ▲ 0.07%COIN 110.51 ▼ 3.92%MSTR 90.32 ▼ 6.40%