Google has released Gemini 4 Argon, an AI model aimed at tasks like coding, research, writing, and especially defensive cybersecurity. Google says Argon can autonomously find, validate, and patch critical software vulnerabilities. The model is being rolled out only to a select set of cyber partners via its Fairwind Program, and Google positions it as outperforming other recent models on Vals benchmarks.
Google DeepMind announced Gemini 4 Argon as the first Gemini 4-generation model aimed at long-horizon software engineering, enterprise knowledge work, and cyber defense. It can generate up to 1M output tokens in a single response, up from 64K on earlier Gemini models. Output-length and safety-guardrail changes enable larger refactors and longer reports in one turn, with a phased U.S. government pre-release process and pricing starting at $2 per 1M input tokens and $10 per 1M output tokens.
Google released Gemini 4 Argon as its flagship model and positioned it as leading on many coding, knowledge-work, and cybersecurity benchmarks. Artificial Analysis’s Intelligence Index scores Gemini 4 Argon at 53 points—tying GPT-6 Astra—but below Claude Opus 5.5 at its high setting. The independent results and internal-use complaints shift expectations toward Opus 5.5 for raw performance while highlighting Argon’s lower per-token and per-task price.
Google unveiled Gemini 4 Argon, positioning it against Anthropic and OpenAI and saying it is already using the model internally. Argon’s single-response output limit rose to up to 1,000,000 tokens from 64,000, alongside a launch price of $2 per million input tokens and $10 per million output tokens. The model is being released first to selected cybersecurity experts and then to paying API customers and Google AI Ultra subscribers, with broader public access left without a set date.
Google announced Gemini 4 Argon, an AI model it says is strongest for coding, knowledge work, and cybersecurity, but it is not available for public use yet. Google says Argon saves 300 TiB of memory across its data centers and reports 77.9 percent on the DeepSWE v1.1 benchmark. Access is limited to Google engineers for now, while internal use and Rust-migration/benchmarking efforts continue until broader availability.
Google announced Gemini 4 Argon, a new frontier model rolling out first to a set of trusted cyber defenders via its Fairwind Program. It will launch with an introductory price of $2 per million input tokens and $10 per million output tokens. The release is staged with guardrails, government pre-release access, a larger 1M-token output limit, and gradual expansion to paid API customers and Google AI Ultra subscribers.
Google announced Gemini 4 Argon as its flagship model but is limiting access while it gathers early feedback and adjusts guardrails. The model is priced at $2 per million input tokens and $10 per million output tokens during an introductory period, before rising to $4 and $20. Access will roll out first to Fairwind participants, then paid API and AI Ultra subscribers, and only later to developers, enterprises, and consumers; overall it performs strongly on knowledge-work benchmarks but is weaker on some coding and cybersecurity measures.
Real World Voice EQ Bench introduced a multidimensional benchmark to evaluate voice AI in real conversational conditions across ASR, speech understanding, STS, and text-to-speech.
NVIDIA, MIT, and Oxford introduced Physis-Lang, a framework that uses physics-focused captions as a shared representation for video world models. On the Physics-IQ Verified leaderboard snapshot dated September 29, 2026, Cosmos3-Super ranked first with 48.2 ± 1.4, beating Veo 3.1. The approach adds physics reasoning and negative prompts, then uses a caption-evolution loop and targeted data curation to improve physical correctness.
Open TTS Leaderboard introduces objective-metric evaluation to rank multilingual text-to-speech and voice cloning models instead of relying on human-vote arenas. It reports macro-average English WER using Seed TTS Eval and CV3 Eval, as of Sep 30, 2026. It adds faster scoring (hours vs weeks) plus optional voice-cloning and streaming metrics (TTFA) alongside an interactive listen-and-vote tab for human preference.
Every AI story that matters,
in your inbox by 8am.
TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the
day in two minutes. Follow companies and topics for alerts, or get the
briefing in Slack. Free, no spam, unsubscribe anytime.
Reading TLDRocket needs no cookies, and the readership counts we rely on come from
our own cookieless analytics. Google Analytics is the exception: it sets cookies and
reports to Google, so it stays switched off until you allow it. You can change your
mind any time from “Cookie settings” in the footer.
Strictly necessary
Session security and form protection (tldrocket-session,
XSRF-TOKEN, 2 hours). The site cannot work without them,
so they need no consent.
Always on
Google Analytics 4 (_ga,
_ga_<id>, up to 2 years). Measures which
stories and sections readers use. Google acts as a third-party processor and may
store the data outside the EU. No advertising, no profiling, no data sold.
