Alphabet (GOOGL) Stock Jumps 2% as Google Unveils Gemini 4 Argon AI Model

Key Highlights

Table of Contents

Google introduced Gemini 4 Argon, its latest AI model, on Wednesday afternoon.
Alphabet shares jumped approximately 2% during Thursday’s premarket session.
Wednesday’s regular trading session saw the stock gain around 1%, with nearly 2% additional growth after hours.
The model’s pricing stands at $2 per million input tokens and $10 per million output tokens, significantly undercutting competitors.
Initial access to Argon is limited to cybersecurity professionals, with broader availability planned for developers and general users.

Alphabet (GOOGL) stock surged approximately 2% during premarket hours on Thursday following Google’s introduction of Gemini 4 Argon, its cutting-edge AI model.

Alphabet Inc., GOOGL

The announcement took place Wednesday afternoon. Alphabet shares had already recorded gains of roughly 1% during regular hours, after reaching session highs near 3% earlier in the day.

CEO Sundar Pichai shared details about Argon through his social media channels. According to Pichai, Google’s internal teams are currently deploying the model for diverse applications including software development and quantum computing research.

Today we’re introducing Gemini 4 Argon.

It delivers frontier performance in complex workflows across real-world software engineering, knowledge work, and cybersecurity defense with an industry-leading 1M token output limit. pic.twitter.com/oDaCoOLu5E

— Google (@Google) September 30, 2026

The company has adopted a staged rollout strategy for Argon. Initial distribution targets cybersecurity specialists and paying API customers exclusively.

Software developers, enterprise clients, and general consumers will receive access in subsequent phases. Google AI Ultra subscription holders are also included in the first deployment group.

Aggressive Pricing Strategy Challenges Competition

Argon’s cost structure represents a significant competitive advantage. Google established pricing at $2 per million input tokens and $10 per million output tokens.

This represents substantial savings compared to Anthropic’s Claude Opus 5.5, which charges $4 and $20 respectively. Claude Fable 5.1 commands even higher rates at $10 and $50.

Additionally, Google provides a 95% discount on cached input tokens. The duration of this promotional pricing structure remains unannounced.

According to Google’s benchmark data, Argon achieved a 77.9% score on the DeepSWE v1.1 coding assessment, setting a new record. However, Jefferies analysts observed that it still lags behind leading models on the more challenging FrontierSWE v2 evaluation.

The model secured first place for coding performance on LLM Arena’s Text Arena. On Code Arena: Web Dev, it claimed eighth position—a remarkable improvement of 21 positions compared to the previous Gemini 3.8 Flash iteration.

Cybersecurity Capabilities Take Center Stage

Google designed Argon specifically with cybersecurity applications as a priority. The company claims the model possesses autonomous capabilities to identify, validate, and remediate software security flaws.

According to Jefferies, Argon matched Grok 4.7’s leading score of 68% on CWE-bench v1, an evaluation measuring vulnerability remediation effectiveness. Cybersecurity company Wiz has begun implementing Argon within its Scan for Good initiative.

Google also documented a reduced failure rate for prompt-injection attacks with Argon at 0.7%. This compares favorably to the 1% rate observed for both Claude Opus 5.5 and Fable 5.1, according to Jefferies data.

The model’s output capacity received a substantial upgrade. Argon now supports up to 1 million tokens, a dramatic increase from 64,000, substantially exceeding Anthropic’s 128,000 token limitation.

Jefferies maintains a “buy” rating on Alphabet with a $445 price target. The firm highlighted that shares had declined 16% from their May 18 high amid concerns that Google was losing ground in the AI competition.

The research firm emphasized that practical deployment will be necessary to validate these initial benchmark results. Neither OpenAI nor Anthropic provided responses to inquiries regarding Argon’s performance claims.

This launch arrives after a challenging period for Google’s DeepMind AI division. Delayed model releases, departures of prominent researchers, and underwhelming benchmark performance had generated investor anxiety prior to Wednesday’s reveal.

Pichai participated in a White House conference on AI safety coinciding with the launch date. He joined fellow technology executives in endorsing a declaration supporting voluntary internal safety protocols and independent audits of AI systems.

Google indicated plans to strengthen protective measures before making Argon publicly available. The company intends to track the model for potential misuse throughout the current limited distribution phase.

✨ Limited Time Offer