Gemini 4 Argon climbs to #5 on Arena's agent board, Opus 5.5 holds #1

Gemini 4 Argon is climbing. On Arena's agent leaderboard, Google's newest frontier model now sits 5th. Ahead of it are Claude Opus 5.5, GPT-6.1 Sol, Claude Fable 5.1 and GPT-6 Astra. On 1 October it was 8th, from 3,417 sessions and with a wide rank range of 3rd to 16th. More votes have pushed it up, but it is still behind its rivals.

The agent board is a different picture from Arena's text leaderboard. There Argon debuted at #1 with 1525 Elo, though that rating is still marked preliminary. Independent benchmarks are just as mixed. Argon tops the Vals Index (68.9% vs Opus 5.5's 67.0%). On the Artificial Analysis Intelligence Index it ties GPT-6 Astra and Fable 5.1 at 53, behind Opus 5.5 at 58. It also uses about 62K output tokens per task, against Astra's 27K.

Price is where Argon competes hardest. It became broadly available on 13 Oct after a cyber-defenders-only preview. Introductory pricing is $2/$10 per 1M tokens until 30 Nov, then it doubles to $4/$20.

Why it matters: Argon wins on chat and price, but on agentic work, where the money is, Anthropic and OpenAI still hold the top four spots.