# OpenAI’s GPT‑6.2 Astra and Sol rank #2 in Intelligence Index

> The new models cost less per token and lead several performance charts, trailing only Anthropic’s Claude Fable 5.2.

Oossa · 2026-11-17 · https://oossa.com/en/openai-s-gpt-6-2-astra-and-sol-rank-2-in-intelligence-index

OpenAI launched two new language models, GPT‑6.2 Astra and GPT‑6.2 Sol, on November 17, 2026. In the Artificial Analysis Intelligence Index, Astra lands at rank 2 with a score of 62, and Sol follows at 57. The move pushes OpenAI just behind Anthropic’s Claude Fable 5.2, which holds the top spot.

Both models are priced to be cheaper per token. Astra costs $8 for input and $40 for output per million tokens, a 20 % drop from the previous GPT‑6 Astra. Cached input for Astra is now $0.40 per million tokens, down from $1.00. Sol keeps the same rates as its predecessor, $2 for input and $10 for output per million tokens.

## How the models compare

Astra climbs five points over GPT‑6 Astra, overtaking Google’s Gemini 4 Carbon (score 60) and Anthropic’s Claude Opus 5.5 (score 59) at their highest effort settings. Sol gains two points, tying GPT‑6 Astra at roughly 15 % of the cost‑per‑task frontier.

OpenAI now occupies most of the Cost‑per‑Task Pareto frontier – a set of models that give the best cost for a given performance – alongside only Anthropic and Meta. It also dominates the Output‑Token‑per‑Task frontier, with Claude Fable 5.2 the sole non‑OpenAI model on that line.

## Strengths and weak spots

The new models shine in science‑focused tests. Astra leads the Terminal‑Bench‑Science 0.1 benchmark for agentic research workflows and ranks second on physics reasoning (CritPt v2 hard subset). In factual knowledge (AA‑Omniscience v1.1 Accuracy) Astra trails only Claude Fable 5.2.

In 3‑D creation, Astra tops the AA‑3D benchmark with a score of 1,318, ahead of Claude Fable 5.2 (1,271) and Sol (1,259). Both Astra and Sol also lead the AA‑Robotics v0.1 preview, scoring 79 and 78 versus Claude Fable 5.2’s 75.

Coding ability did not improve. Astra matches GPT‑6 Astra, while Sol drops three points on the Coding Agent Index, mainly because it omits key facts more often on the SWE‑Atlas‑QnA‑AA test.

## The facts

- OpenAI released GPT‑6.2 Astra and GPT‑6.2 Sol on 2026‑11‑17.
- Astra’s price is $8/$40 per million input/output tokens, 20 % lower than GPT‑6 Astra.
- Cached input for Astra is $0.40 per million tokens, down from $1.00.
- In the Artificial Analysis Intelligence Index, Astra scores 62 (rank 2) and Sol scores 57.
- Astra leads the AA‑3D benchmark with a score of 1,318.

## Why it matters

For developers, the lower token prices mean cheaper API calls for large‑scale projects. The higher science and 3‑D scores suggest these models can handle more complex research and simulation tasks better than most competitors, though their coding performance has not improved.

## Sources & references

1. [OpenAI has the #2 model in the Intelligence Index with the release of GPT-6.2 Astra and GPT-6.2 Sol](https://artificialanalysis.ai/articles/openai-has-the-2-model-in-the-intelligence-index-with-the-release-of-gpt-6-2-astra-and-gpt-6-2-sol) – Artificial Analysis, 2026-11-17
2. [[AINews] GPT-6.2 Astra, the #2 Frontier Model is better and 20%-600% cheaper, but has mixed reviews; 6.2 Sol likely main workhorse](https://latent.space/p/ainews-gpt-6-2-astra-the-2-frontier-model-is-better-and-20-600-cheaper-but-has-mixed-reviews-6-2-sol/) – Latent Space, 2026-11-18
3. [GPT‑6.2 Astra and Sol](https://openai.com/index/introducing-gpt-6-2-astra-sol/) – Hacker News, 2026-11-17
4. [OpenAI launches GPT-6.2 Astra and Sol with lower API prices and some benchmark gains](https://the-decoder.com/openai-launches-gpt-6-2-astra-and-sol-with-lower-api-prices-and-some-benchmark-gains/) – The Decoder, 2026-11-17
5. [OpenAI Releases GPT-6.2 Astra and Sol With a 20% Cheaper Astra API](https://www.marktechpost.com/2026/11/17/openai-gpt-6-2-astra-sol-release/) – MarkTechPost, 2026-11-17

Last updated: 2026-11-18
