Google is doubling the price of its flagship Gemini 4 Argon model on the Gemini API. From December 1, the standard rate rises from $2 to $4 per million input tokens and from $10 to $20 per million output tokens, according to Google’s own pricing page. Cached input doubles too, from $0.10 to $0.20, and the Batch, Flex and Priority tiers all rise by the same proportion. The change lands four weeks after Argon launched on October 13 at an introductory rate, and it leaves Google’s flagship twice as expensive as OpenAI’s GPT-6.1 Sol.
What changes on December 1
The rise covers every way of calling Argon. On the Standard tier, input tokens go from $2 to $4 per million and output tokens — including the thinking tokens the model spends reasoning — from $10 to $20. Cached input, which Google prices at 95% off the input rate, doubles from $0.10 to $0.20 per million.
The cheaper tiers move by the same proportion. The Batch API, which halves the standard rate for work that can wait, rises from $1 and $5 to $2 and $10 per million tokens. Flex pricing follows the same path, and the Priority tier for production workloads doubles from $3.60 and $18 to $7.20 and $36.
The field has moved the other way
Argon’s rise lands in a month when rival prices have mostly fallen. OpenAI lists GPT-6.1 Sol at $2 per million input tokens and $10 per million output tokens, the same rates Argon charged until now. OpenAI cut Sol’s price in half on November 2, from $4 and $20 against GPT-5.6 promotional pricing.
Anthropic’s Claude Opus 5.5 lists at $4 and $20, where Argon lands in December, while DeepSeek’s V4.1 Flash sits far below all three at $0.30 and $1.20 per million tokens at peak rates. API price rises are not new this month — Together raised Qwen3.7-Max prices by two-thirds — but Google’s move is the first doubling by a frontier lab.
What to do before December 1
Teams running Argon in production have three weeks to decide. The cheapest lever is caching: prompts served from cache stay at $0.20 per million after December, a twentieth of the fresh-input rate, so workloads with repeated system prompts barely feel the rise. Batch processing halves every rate, which suits overnight pipelines and evaluation runs that can wait a day.
The harder question is whether Argon is still worth twice Sol’s price. Both models now sit at opposite ends of the same $2-to-$4 range, and the honest test is a side-by-side run on the team’s own prompts, not a benchmark table.
Google’s pricing page already prints the December rates, so there is no grace period beyond November 30. Teams that wait until December to check their usage will find the higher rates already applied to the first invoice of the month.
Why It Matters
The price rise is a bet that Argon’s quality now justifies a premium. Google launched the model as the cheapest way into frontier reasoning, matching Sol dollar for dollar, and spent four weeks building its case — top of the Vals Index, wins in Chrome refactoring and fusion research, and a robotics variant.
From December, developers pay double for that case, and the choice becomes explicit: stay on Argon, or re-route the same workload to Sol at half the price. Batch pricing softens the blow for overnight jobs, but it doubles too, so the saving is relative, not absolute.
The teams most exposed built on Argon during its cheap weeks and now face a bill that doubles with no change in quality. Google’s gamble is that the model’s results — and the switching cost of re-tuning prompts and tools — keep them paying. If workloads move to Sol in December, the price war’s next move belongs to Google.
Background
Argon arrived on October 13 as Google’s latest frontier model, topping the Vals Index on launch and shipping across the Gemini API, AI Studio, Google Cloud and Antigravity. Its sibling, Gemini 4 Carbon, followed on October 28 with a cybersecurity focus, and Google made its video model, Gemini Omni 1.2 Flash, generally available on November 5.
Source: Google’s Gemini API pricing and the Gemini 4 Argon launch post.
0 Comments