Gemini 4 Argon by @GoogleDeepMind just went head-to-head with top frontier models across a set of generations. At 33% lower cost per task than GPT-6.1 Sol, and 70% lower than Claude Opus 5.5, see first impressions from @petergostev on how Gemini 4 Argon compares. piped.video/watch?v=h5EL5zTh…

Gemini 4 Argon | First Impressions

Try the models yourself: https://arena.ai/agentGoogle just droppe...

youtube.com

Oct 1, 2026 · 12:09 AM UTC

26
15
308
31,716
Sort replies: Relevant Recent Liked
Replying to @arena @GoogleDeepMind @petergostev
There is a lot of controversy surrounding this clip... Could you please provide more diverse tests?
137
Replying to @arena @GoogleDeepMind @petergostev
33% cheaper. Does the unit still clear?
113
Replying to @arena @GoogleDeepMind @petergostev
33% cheaper gets Argon on the shortlist. show me the generation logs.
535
Replying to @arena @GoogleDeepMind @petergostev
as succeeder of 3.8 flash, it is good, but as flagship model of google, it is failure. so i prefer to call it as "gemini 4.0 flash"
1
256
Replying to @arena @GoogleDeepMind @petergostev
What's the point of assessing text-generating models on graphic design?
68
Replying to @arena @GoogleDeepMind @petergostev
Gemini 4 Argon is over hyped in my opinion. It's a good model but not the best. It is hyped like the best which just makes it worse when we use it and find out it isn't
1
271
Replying to @arena @GoogleDeepMind @petergostev
Does cost per task include thinking tokens? With output allowed up to 1M, I'd expect Argon's per-task cost to swing a lot depending on how long it decides to reason.
78
Replying to @arena @GoogleDeepMind @petergostev
We need to compare all previous Gemini pro and flash model to Gemini 4 Argon.
128
Replying to @arena @GoogleDeepMind @petergostev
Gemini 4 Argon undercutting the big names on cost while still hanging with them on quality is a pretty wild first look.
194
Replying to @arena @GoogleDeepMind @petergostev
Gemini 4 Argon battles frontier models, costing 33% less than GPT-6.1 Sol and 70% less than Claude Opus
47
Replying to @arena @GoogleDeepMind @petergostev
hello, please bring back sonnet 4.6 🙏🙏🙏🙏
18
Replying to @arena @GoogleDeepMind @petergostev
hello please bring back sonnet 4.6
76
Replying to @arena @GoogleDeepMind @petergostev
Make it included in the pro plan pls
7
Replying to @arena @GoogleDeepMind @petergostev
how does the reasoning hold up on complex multi-step tasks?
451
Replying to @arena @GoogleDeepMind @petergostev
Top tier results at fraction of cost pure efficiency win
120
Replying to @arena @GoogleDeepMind @petergostev
Wow, very fast!
326
Replying to @arena @GoogleDeepMind @petergostev
solid launch
399
Replying to @arena @GoogleDeepMind @petergostev
Google cut cost by 30 percent against Sol and seventy percent versus Claude Opus Five Point Five for the same generation task
253
Replying to @arena @GoogleDeepMind @petergostev
33% under sol, 70% under opus. the discount leaderboard moves fast
387