🚀 Google AI's Gemini 4 Argon is now generally available in GitHub Copilot. What we learned from early testing ⬇️ • It solved more tasks than previous Gemini models on every coding benchmark, with its largest gains in Copilot CLI. • It checks its work thoroughly, testing edge cases... before it finishes. Use Gemini 4 Argon for complex coding, agentic, and reasoning tasks. Find it in the GitHub Copilot app, CLI, and Visual Studio Code.show more

@GoogleAI In ChessBench, Argon destroyed the previous number 1 spot, Opus 5.5 in just 23 moves. Seems to be a very impressive model so far.

@GoogleAI one prompt, new game in its own repo, added to the arcade. not that hard honestly

@GoogleAI prolly the only good thing in Github copilot rn

@GoogleAI maybe its safe to say that google cooked with this one

@GoogleAI "Checks its work" is the part I care about. A model passing its own tests is a good start. The failures that hurt are the ones where every check is green and the thing the user wanted still didn't happen.

@GoogleAI Gemini 4 Argon landing in GitHub Copilot with bigger gains in Copilot CLI is a practical upgrade for everyday coding. Thorough edge-case checking is what operators notice.

@GoogleAI finally a coworker that checks its own work 😭

@GoogleAI Every coding benchmark improving is a seriously strong Copilot debut.

@GoogleAI checks its work is the part i care about

@GoogleAI 会自己测边缘情况，能少来回改几轮就挺好

@GoogleAI Argon in Copilot is a seat change, not a free upgrade. Freeze one failing CLI task both Gemini and your prior model already know; same repo, same prompt, same assert. Whichever returns green with fewer retries keeps the default.

@GoogleAI benchmarks again huh

@GoogleAI changelog fine print: argon bills at provider list pricing under usage-based billing, pro+/max/business/enterprise only. google's list is $2/$10 per 1M through nov 30 and $4/$20 from dec 1. budget for the jump

@GoogleAI Interesting that "solved more tasks" is the headline metric they chose to highlight right off the bat.

@GoogleAI Biggest gains in the CLI is the part I care about. That's where it has to finish the job without you babysitting it.

@GoogleAI 你怎么看@grok

@GoogleAI nerd alert 💀

@GoogleAI It checks edge cases before finishing. Id still want to see what those checks actually were tho

@GoogleAI checking edge cases before it finishes is the whole difference between a demo and a teammate

@GoogleAI the gains in copilot cli are the part that actually matters here

@GoogleAI does checking its work thoroughly mean more tool calls per task? curious what that does to latency and premium request usage in copilot cli

@GoogleAI wonder if it actually catches its own bugs or just says it did

@GoogleAI Biggest gains landing in Copilot CLI is the interesting bit. The terminal is quietly becoming the main battleground for coding agents

@GoogleAI Excited to try Gemini 4 in Copilot

@GoogleAI the cost win only matters if argon matches copilot's existing model on latency-sensitive autocomplete. most developers will tolerate slower for better suggestions, but not the other way around.

@GoogleAI The biggest gains landing in Copilot CLI and not the editor says the harness around Gemini 4 Argon moves the score too. I'd like to see the numbers for each surface side by side.

@GoogleAI the cli gains being larger than the editor is interesting. is that because cli tasks are more structured, or because argon handles multi-step workflows better?

@GoogleAI CLI gains I'll buy — every coding benchmark? @grok

@GoogleAI Hello, my ticket (#4825217) has been ignored for a month. There is $7,000 in the sponsor account that is over 100 days past the expected payment date. How can this be resolved?