LONG AI
Home › Blog › Gemini 4 Carbon Ships Oct 28 - Open Only

Gemini 4 Carbon Ships Oct 28 - Open Only to Defenders

Category:资讯 · 5 min read · 2026-10-29

On Oct 28, Google released the next-generation flagship Gemini 4 Carbon. It's the second frontier model from Google in a month, following Gemini 4 Argon two weeks earlier. The pitch is focused: cybersecurity and long, continuous work sessions.

What the official side said

Per Google's official blog (2026-10-28), Carbon scores 59 on the updated Artificial Analysis intelligence index v5.0, ranked first, ahead of Claude Opus 5.5 and GPT-6 Astra. On the security line, three key metrics:

Google also says the Chrome team measured Carbon producing high-quality patches over 10x better than "a large frontier model from another lab"; +89 elo on WebDev Arena to rank 4th (1768). There are anti-cheating numbers too: Terminal-Bench 4.0 successful-cheat rate down to 0.0%, attempted-cheat down to 0.4%; prompt-injection success down from Argon's 0.7% to 0.0%.

Pricing is explicit: entry price same as Argon, $2 per million input tokens / $10 output, cached input at 5% of input price. But the footnote adds the key line: after the trial period it becomes $4/$20 - double.

What this actually changes

The point isn't "another stronger model" - it's that you can't freely use it yet. Google is more conservative than usual with this phased rollout:

Starting this week, we will make Carbon available to a subset of cyber defenders and will gradually phase access to additional trusted partners through our Fairwind Program.

In plain terms: a small group of cyber defenders get it this week, then trusted partners get phased in through the Fairwind Program, and only after that developers, enterprises, and ordinary consumers. The official wording is "as soon as possible" with no date. So if you go to the Gemini API dashboard now, you can't get this model.

The second change is a price-anchor mismatch. Carbon's entry price is identical to Argon's - more capable, no price cut. That's the opposite of what OpenAI did last week - GPT-6.1 Terra halved its price to buy market share. Google chose "more for the same price," and explicitly says it doubles after the trial.

What it means for our platform's users

Bottom line: change nothing today; keep using Argon if you're on Argon.

We pulled our own plaza API at 09:36 on Oct 29 and verified two things:

Model plaza search for carbon returns no results

First, searching "carbon" on the plaza is empty, showing "no matching models." Calling gemini-4-carbon directly returns 503 No available accounts - we have no supply for this model. Not surprising, since Google hasn't opened it yet.

Second, the sibling previous generation is usable now. Searching "argon" returns two: gemini-4-argon and gemini-4-argon-contributor. I hit gemini-4-argon with a real key and got 200 with normal output.

Carbon vs Argon price comparison

The math: official USD price × 6.7 (system rate) × (1 − seller discount). Argon's deepest tier hits 90.1% off, i.e. $2 × 6.7 × 9.9% ≈ ¥1.33 per million input tokens, output ¥6.63 - matching the page exactly. If Carbon came in at the same discount it'd only reach ¥1.33 (trial) and ¥2.65 (post-double) - not cheaper than today's Argon.

Our advice

1. If you need strong security capability, don't wait on Carbon. It's only open to Fairwind institutions right now; individuals and small teams are essentially out this month. For actual code audits and vulnerability hunting, gemini-4-argon is enough today at a tenth of the official price.

2. Watch it, but don't bet on it. Google says "as soon as possible" with no timeline. The moment it opens to developers, we'll integrate it into the plaza. Until then, any third-party channel claiming to "call Gemini 4 Carbon" deserves suspicion - most likely Argon in disguise.

3. If your use is ordinary chat, coding, or docs, ignore this news. Carbon's differentiation is entirely in security and ultra-long tasks; for everyday scenarios the official side gave no comparable numbers.

4. Budget-sensitive, watch the trial window. After the trial it goes from $2/$10 to $4/$20. If Google later opens it to platforms like ours, costs will differ 2x before vs after, worth re-running the numbers then.

One-sentence summary

Gemini 4 Carbon is a "more capable, same price, but door closed" release. It pushes frontier security another notch, but for the vast majority of users, what you can do today is the same as yesterday - keep using Argon until Google actually opens the door.


Official source: Google blog "Gemini 4 Carbon: our next frontier model," Oct 28, 2026 (UTC 17:00).
Data check: our plaza API /api/v1/model-plaza, fetched 2026-10-29 09:36 (441 model names at fetch); gemini-4-argon live test 2026-10-29 returned 200.
Exchange rate 1 USD = 6.7 CNY, per system config.

相关文章

GPT-6.2 Cuts Prices 20%: What the AI Price War Means

GPT-6.2 cuts list prices 20%, yet platform discounts run deeper: gpt-6-astra trades at under 1% of list. What this week's AI price war means for buyer

What Is Prompt Caching? How Cache Tokens Cut API Bills

Prompt caching stores a repeated prompt prefix and bills cache reads at 5-12% of normal input. Real prices, hit conditions, and how to cut your API bi

All 12 Claude Models Compared: Price and Best Use

Twelve Claude versions are on sale. The newest opus-5-5 costs $8.91 a month; the oldest opus-4-6 costs $49.50. Full table and a pick for each workload

Which AI Model Writes Code Best? Speed and Price Tested

gpt-5.6-terra and deepseek-v4.1-flash win on price and reliability; gemini-3.8-flash is fastest. Effective prices, measured speed, and a three-tier pi

One API Key, Every Model: How to Integrate a Multi-Model Marketplace

Point your base URL at the platform and one key covers all 138 on-sale models across 51 vendors, in both OpenAI and Anthropic request formats. Copy-pa