TLDRocket
Sign in

Gemini 4 Matches GPT-6 Astra but Trails Opus 5.5

Trending Topics Jakob Steinschaden ● Covered by 7 sources

Google’s new Gemini 4 Argon lands level with OpenAI’s GPT-6 Astra, but behind Anthropic’s best. The twist: it may be the cheap one, not the strongest one.

Based on reporting by Trending Topics, Jakob Steinschaden — read the original for the full story.

Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error

Google has a new flagship model after almost a year, and it came out swinging. Gemini 4 Argon is being pitched as a leap for coding, knowledge work, and cybersecurity defense, with support for responses as long as one million tokens. In Google’s own table, it led 13 of 19 benchmarks. But the independent view is less dramatic.

Artificial Analysis puts Argon at 53 points on its Intelligence Index, which leaves it essentially tied with OpenAI’s GPT-6 Astra and clearly behind Anthropic’s Claude Opus 5.5. On the tested High compute setting, Argon scored 52.6 points. Astra came in at 52.7. Opus 5.5 reached 57.6, and Anthropic’s smaller Sonnet 5.5 also stayed ahead at 56. Google ended up in eighth place overall, with the top seven spots occupied by Anthropic variants and Astra.

The individual tests tell the same story, just with a few wrinkles. Argon beat Astra on Humanity’s Last Exam and on GDPval-AA, the work-style test built from 44 occupations. Astra, meanwhile, was slightly ahead on Terminal-Bench 4.0, and Opus 5.5 stayed in front there too. On SciCode, the gap widened again in Anthropic’s favor. One detail stands out: Google’s own Terminal-Bench number for Argon, 57.4 percent, was almost identical to Artificial Analysis’s 57.1 percent.

Then there’s price, where Google looks much better. Argon is listed at $2 per million input tokens and $10 per million output tokens, half the price of Opus 5.5 and a fifth of GPT-6 Astra on paper. Artificial Analysis also estimates the cost of an average task, and Argon comes in at $1.99, versus $3.26 for Astra and $5.98 for Opus 5.5. Google is also offering Argon at an introductory price, but hasn’t said how long that will last.

That cheapness matters because Google’s model is not broadly available yet. For now it’s only going to selected cybersecurity teams, with paying API customers and Google AI Ultra subscribers coming later. And yes, that’s the real story here: not who can write the loudest launch post, but who can ship a model that’s close enough on performance and a lot less painful on the bill.

My take — AI-written commentary, not fact-checked reporting

This is the AI market in one paragraph: the grandest model isn’t always the smartest buy, and the smartest buy isn’t always the winner on the chart. Google can call Argon frontier intelligence all it wants; the real flex is pricing it like a lure and hoping nobody notices the introductory sign. The industry loves benchmark trophies, but customers still pay invoices.

Read more about this at: Trending Topics

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.