Market Flux Event
Google Launches Gemini 4 Argon with 1M Output Tokens and Cybersecurity Focus as OpenAI Partners with Synopsys on Chip Design AI
Read this in the Market Flux appGoogle unveiled Gemini 4 Argon on September 30, its most advanced AI model and the flagship of the Gemini 4 generation, after months of delays following strong pretraining evaluations reported as early as September 1. Argon is priced at $2 per million input tokens and $10 per million output tokens, with cached input priced 95% lower, and dramatically expands the output token limit to 1 million tokens per trajectory, up from a previous ceiling of 64,000.
Google positions Argon as competitive with OpenAI's Astra and Anthropic's Opus 5.5 across coding, enterprise work, and cybersecurity. The model leads several published benchmarks, including a 77.9% score on DeepSWE v1.1, 91.7% on LVBench, 68.9% on the Vals Index, 51.3% on AutomationBench, and 65.4% on Finance Agent v2. Argon also tied for the top score of 68% on CWE-bench v1, a cybersecurity vulnerability benchmark.
Argon is already in active use inside Google by thousands of employees. Google says Argon agents have identified memory optimizations freeing more than 300 TiB across its data centers, with an estimated 500 TiB to 1 PiB in total projected savings. The model also helped optimize a quantum computing subroutine by 40% and is being used for large-scale C/C++ to Rust code migrations, including more than 800,000 lines of the Fuchsia Zircon kernel.
Cybersecurity is a central capability. Google says Argon can autonomously find, validate, and patch software vulnerabilities, and will release the model to trusted cyber defenders without its standard guardrails so they can deploy its full defensive power. Google is deploying misalignment mitigations that monitor Argon's chain-of-thought and halt execution when necessary. Broader rollout to paid API customers and Google AI Ultra subscribers will follow early tester feedback, with Google also participating in the Trump administration's voluntary pre-release model access process before wider availability.
Separately, OpenAI and Synopsys announced a partnership on September 30 to develop GPT-Synopsys, a frontier AI model purpose-built for chip design work. The deal links OpenAI's model capabilities with Synopsys's electronic design automation expertise at a moment when the chip design industry faces rising complexity and demand tied to AI infrastructure buildout. OpenAI has also been marketing its UltraFast tier at $500 per month, positioning speed as a premium differentiator in an increasingly competitive frontier model market.
© AI-generated summary is provided by Market Flux
Sources
- YahooFinanceOpenAI is selling speed, and it costs $500 a month. The demos are where UltraFast gets wild. In the full episode, we break down what instant AI can actually do and why the chip race behind it matters. Watch in The Loop on YouTube: 🎙️ @cryptopunk7213
- DeItaone$SNPS - OPENAI AND SYNOPSYS ANNOUNCE GPT-SYNOPSYS: FRONTIER INTELLIGENCE TO REVOLUTIONIZE CHIP DESIGN
- InvestingSynopsys, OpenAI partner on AI model for chip design
- FirstSquawkGOOGLE GRAPPLES WITH EMPLOYEE SKEPTICISM ABOUT THE NEW GEMINI MODEL
- CNBCGoogle rolls out Gemini 4 Argon, its most advanced AI model
- FaststocknewssGOOGLE $GOOGL ANNOUNCES GEMINI 4 ARGON, TRAINED TO BE HIGHLY CAPABLE AT CYBERSECURITY DEFENSE The company says it will release Argon without cyber guardrails to trusted defenders and its own internal teams, and is expanding the model's output token limit to 1M tokens. Google says it is deploying misalignment mitigations that monitor Argon's chain-of-thought and actions and stop execution when necessary. Access is expanding gradually while Google participates in the US government's voluntary process for pre-release model access, with broader availability to developers, enterprises, and consumers after early tester feedback. WSJ reported September 1 that Gemini 4 had performed well on pretraining evaluations but still needed to complete posttraining.
- ReutersGoogle announces Gemini 4 flagship AI model after months of delays
- Wallstengine$GOOGL UNVEILS GEMINI 4 ARGON Google unveiled Gemini 4 Argon, its new frontier model and the top model in the Gemini 4 generation. Here’s everything you need to know: Pricing: $2 per 1M input tokens and $10 per 1M output tokens, with cached input priced 95% lower. Output limit: Argon can generate up to 1 million output tokens in a single trajectory, up from 64K previously. Performance: Google says Argon is competitive with OpenAI’s Astra and Anthropic’s Opus 5.5 across coding, enterprise work and cybersecurity. It leads several published benchmarks, including: Vals Index: 68.9% AutomationBench: 51.3% Finance Agent v2: 65.4% DeepSWE v1.1: 77.9% LVBench: 91.7% Internal use at Google: Argon is already being used by thousands of employees. Google says Argon agents identified memory optimizations that freed more than 300 TiB across its data centers, with an estimated 500 TiB to 1 PiB of total savings. It also helped optimize a quantum computing subroutine by 40%, and is being used for large-scale C/C++ to Rust migrations, including 800K+ lines for the Fuchsia Zircon kernel. Cybersecurity: Argon scored 68% on CWE-bench v1, tying for the top score, and Google says it can autonomously find, validate and patch software vulnerabilities. Trusted cybersecurity partners will get access without the model’s normal cyber guardrails so they can use its full defensive capabilities. Availability: Argon is initially rolling out to select cyber defenders and Google’s internal teams. Google says broader access will come later, starting with paid API customers and Google AI Ultra subscribers, but no public release date was given. Google is also participating in the Trump administration’s voluntary pre-release model access process before wider availability