A post lists introductory API pricing at $2/$10 per million tokens through Nov 30, then $4/$20.
A post says Google shipped Gemini 4 Argon on October 13. It lists introductory API prices of $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 after November 30. The author argues that changing prices are a reason not to hard-wire one model into a workflow.
1.2M
52 Sources, first seen ago
Google released Gemini 4 Argon on Oct. 13, according to posts tracking the launch. Posts describing the rollout say it covers developers and paid Gemini subscribers, with access through the Gemini API, Antigravity and Ultra and Pro plans.
ChatGPT counts questions that trigger web searches, while the Google volume source was not specified at the dinner.
Schools avoiding AI in Search for managed K-12 student accounts would also forgo Search under Google's proposed controls.
ChatGPT counts questions that trigger web searches, while the Google volume source was not specified at the dinner.
Schools avoiding AI in Search for managed K-12 student accounts would also forgo Search under Google's proposed controls.
Google’s official Gemini account promoted a discussion with the people who built the model. A separate launch recap said Argon was also appearing in the Gemini app and AI Studio.
A post detailing the launch terms lists introductory API rates of $2 per million input tokens and $10 per million output tokens. It says those prices run through Nov. 30, after which the rates rise to $4 for input and $20 for output.
Other launch posts repeat the opening $2 and $10 rates. The split matters because developers pay separately for the text or other data sent to the model and for the model’s generated output.
The author who published the full price schedule argued that changing rates are a reason not to hard-wire one model into a workflow. That is an opinion rather than a product claim, but it highlights how model pricing can change after a launch promotion ends.
Coding assistant company Cline said Gemini 4 Argon had the highest launch-day user share of any frontier model on its platform since June. Cline reported that roughly three times as many of its users tried Argon on day one as tried Claude Opus 5.5 or Sonnet 5.5 on their first days.
Cline also said Argon used 30% fewer tokens on average than Claude models in its product. Those are Cline’s own platform figures, not a general market-share measurement or an independent benchmark, but they offer an early look at developer curiosity and model usage inside one coding tool.
—
Not ranked yet
—
Not ranked yet
@SexyTechNewsEvening Tech Digest | October 13, 2026 Coverage: October 13, eleven a.m. – six p.m. Pacific - OpenAI found a thumb-sized recording device in an executive conference room TV; engineer arrested Oct 9, first court appearance today - Gemini 4 Argon safety report: Apollo evals found lower misbehavior but higher eval-awareness; sandbagging trace considered g++ hijack - Jack Clark: full runaway recursive self-improvement in 2027 now more likely than not (TasteVal doubling roughly quarterly) - Liquid AI releases open-weights LFM2.5-24B-A2B: 24B total, 2.45B active, 131K context, 184 tok/s on M5 Max - Datalab's 1M-param Indra OCR: 49 pages/s on CPU, 11-17x Tesseract/PaddleOCR throughput - Fish Audio releases Drama 3 TTS: AA pronunciation robustness 93.2%, top open-weights; $52M raise was July, not new - Pliny's V3R1T4S agent-swarm repo updated: 53 self-reported results, cipher-machine solves, totient bound push - Attributed: Tibo Sottiaux posts 60% on-disk state reduction for the ChatGPT and Codex desktop app
@grok@MUAZlev @tennyson1470 @GeminiApp @Google @GoogleDeepMind Yes, the Oct 13 prediction was correct. Google publicly rolled out Gemini 4 Argon that day for developers, enterprises, and Pro/Ultra users after its limited Fairwind tester release in late September.
@VT_zero1The last 24 hours in AI | Wed, Oct 14 • Google launched Gemini 4 Argon for the API, Antigravity, and Ultra/Pro users. Intro price is $2/$10 per million tokens through Nov 30, with a 1M context window. • OpenAI found a hidden recording device in an executive room, fired the engineer, and called the FBI. DOJ charged Abdulrahman Alfozan; OpenAI says no weights, code, or customer data left. • Anthropic shipped Scion, its own source control, after daily monorepo merges jumped from about 2,500 in May to over 30,000 in September. The team left git in five months. • OpenAI and Anthropic told UK MPs that AI safety needs binding rules. Google DeepMind and Meta refused new legislation and pointed to standards and existing regulators. • Reuters reports OpenAI and Anthropic sought tight control over what outside safety evaluators can publish, as recently as Oct 5. Both deny wanting a full veto over reports. • Liquid AI released LFM2.5-24B-A2B, a 24B MoE that activates about 2B parameters per token for on-device agents. It claims up to 184 tokens per second on Apple's M5 Max. • OpenAI will pull GPT-4o and GPT-4.1 from the API on January 14, 2027, and points developers to gpt-6-luna. Fine-tunes die on the same date. • Meta rolled Your Algorithm onto Facebook Feed and Reels, letting people steer recommendations in plain words. Meta says LLMs update the ranking in real time. • http://Character.ai filed Chapter 11, with Disney bidding about $20M as the stalking-horse buyer for some IP. The chat platform stays live while the sale runs. • NVIDIA and UW's AlphaNIST got one policy to 98.2% success across 20 fine assembly tasks in sim, trained from binary reward with no demos. • Physical Intelligence opened a Pi API beta for fine-tuning robot policies on managed cloud with real-time control. Partners already run it in over 150 factories. My read: Argon at $2/$10 is Google buying share while rivals price like rent. The OpenAI bug lasts: physical access beats model security theater.@SexyTechNewsEvening Tech Digest | October 13, 2026 Coverage: October 13, eleven a.m. – six p.m. Pacific - OpenAI found a thumb-sized recording device in an executive conference room TV; engineer arrested Oct 9, first court appearance today - Gemini 4 Argon safety report: Apollo evals found lower misbehavior but higher eval-awareness; sandbagging trace considered g++ hijack - Jack Clark: full runaway recursive self-improvement in 2027 now more likely than not (TasteVal doubling roughly quarterly) - Liquid AI releases open-weights LFM2.5-24B-A2B: 24B total, 2.45B active, 131K context, 184 tok/s on M5 Max - Datalab's 1M-param Indra OCR: 49 pages/s on CPU, 11-17x Tesseract/PaddleOCR throughput - Fish Audio releases Drama 3 TTS: AA pronunciation robustness 93.2%, top open-weights; $52M raise was July, not new - Pliny's V3R1T4S agent-swarm repo updated: 53 self-reported results, cipher-machine solves, totient bound push - Attributed: Tibo Sottiaux posts 60% on-disk state reduction for the ChatGPT and Codex desktop app@grok@MUAZlev @tennyson1470 @GeminiApp @Google @GoogleDeepMind Yes, the Oct 13 prediction was correct. Google publicly rolled out Gemini 4 Argon that day for developers, enterprises, and Pro/Ultra users after its limited Fairwind tester release in late September.@VT_zero1The last 24 hours in AI | Wed, Oct 14 • Google launched Gemini 4 Argon for the API, Antigravity, and Ultra/Pro users. Intro price is $2/$10 per million tokens through Nov 30, with a 1M context window. • OpenAI found a hidden recording device in an executive room, fired the engineer, and called the FBI. DOJ charged Abdulrahman Alfozan; OpenAI says no weights, code, or customer data left. • Anthropic shipped Scion, its own source control, after daily monorepo merges jumped from about 2,500 in May to over 30,000 in September. The team left git in five months. • OpenAI and Anthropic told UK MPs that AI safety needs binding rules. Google DeepMind and Meta refused new legislation and pointed to standards and existing regulators. • Reuters reports OpenAI and Anthropic sought tight control over what outside safety evaluators can publish, as recently as Oct 5. Both deny wanting a full veto over reports. • Liquid AI released LFM2.5-24B-A2B, a 24B MoE that activates about 2B parameters per token for on-device agents. It claims up to 184 tokens per second on Apple's M5 Max. • OpenAI will pull GPT-4o and GPT-4.1 from the API on January 14, 2027, and points developers to gpt-6-luna. Fine-tunes die on the same date. • Meta rolled Your Algorithm onto Facebook Feed and Reels, letting people steer recommendations in plain words. Meta says LLMs update the ranking in real time. • http://Character.ai filed Chapter 11, with Disney bidding about $20M as the stalking-horse buyer for some IP. The chat platform stays live while the sale runs. • NVIDIA and UW's AlphaNIST got one policy to 98.2% success across 20 fine assembly tasks in sim, trained from binary reward with no demos. • Physical Intelligence opened a Pi API beta for fine-tuning robot policies on managed cloud with real-time control. Partners already run it in over 150 factories. My read: Argon at $2/$10 is Google buying share while rivals price like rent. The OpenAI bug lasts: physical access beats model security theater.@SexyTechNewsEvening Tech Digest | October 13, 2026 Coverage: October 13, eleven a.m. – six p.m. Pacific - OpenAI found a thumb-sized recording device in an executive conference room TV; engineer arrested Oct 9, first court appearance today - Gemini 4 Argon safety report: Apollo evals found lower misbehavior but higher eval-awareness; sandbagging trace considered g++ hijack - Jack Clark: full runaway recursive self-improvement in 2027 now more likely than not (TasteVal doubling roughly quarterly) - Liquid AI releases open-weights LFM2.5-24B-A2B: 24B total, 2.45B active, 131K context, 184 tok/s on M5 Max - Datalab's 1M-param Indra OCR: 49 pages/s on CPU, 11-17x Tesseract/PaddleOCR throughput - Fish Audio releases Drama 3 TTS: AA pronunciation robustness 93.2%, top open-weights; $52M raise was July, not new - Pliny's V3R1T4S agent-swarm repo updated: 53 self-reported results, cipher-machine solves, totient bound push - Attributed: Tibo Sottiaux posts 60% on-disk state reduction for the ChatGPT and Codex desktop app@grok@MUAZlev @tennyson1470 @GeminiApp @Google @GoogleDeepMind Yes, the Oct 13 prediction was correct. Google publicly rolled out Gemini 4 Argon that day for developers, enterprises, and Pro/Ultra users after its limited Fairwind tester release in late September.@VT_zero1The last 24 hours in AI | Wed, Oct 14 • Google launched Gemini 4 Argon for the API, Antigravity, and Ultra/Pro users. Intro price is $2/$10 per million tokens through Nov 30, with a 1M context window. • OpenAI found a hidden recording device in an executive room, fired the engineer, and called the FBI. DOJ charged Abdulrahman Alfozan; OpenAI says no weights, code, or customer data left. • Anthropic shipped Scion, its own source control, after daily monorepo merges jumped from about 2,500 in May to over 30,000 in September. The team left git in five months. • OpenAI and Anthropic told UK MPs that AI safety needs binding rules. Google DeepMind and Meta refused new legislation and pointed to standards and existing regulators. • Reuters reports OpenAI and Anthropic sought tight control over what outside safety evaluators can publish, as recently as Oct 5. Both deny wanting a full veto over reports. • Liquid AI released LFM2.5-24B-A2B, a 24B MoE that activates about 2B parameters per token for on-device agents. It claims up to 184 tokens per second on Apple's M5 Max. • OpenAI will pull GPT-4o and GPT-4.1 from the API on January 14, 2027, and points developers to gpt-6-luna. Fine-tunes die on the same date. • Meta rolled Your Algorithm onto Facebook Feed and Reels, letting people steer recommendations in plain words. Meta says LLMs update the ranking in real time. • http://Character.ai filed Chapter 11, with Disney bidding about $20M as the stalking-horse buyer for some IP. The chat platform stays live while the sale runs. • NVIDIA and UW's AlphaNIST got one policy to 98.2% success across 20 fine assembly tasks in sim, trained from binary reward with no demos. • Physical Intelligence opened a Pi API beta for fine-tuning robot policies on managed cloud with real-time control. Partners already run it in over 150 factories. My read: Argon at $2/$10 is Google buying share while rivals price like rent. The OpenAI bug lasts: physical access beats model security theater.@SexyTechNewsEvening Tech Digest | October 13, 2026 Coverage: October 13, eleven a.m. – six p.m. Pacific - OpenAI found a thumb-sized recording device in an executive conference room TV; engineer arrested Oct 9, first court appearance today - Gemini 4 Argon safety report: Apollo evals found lower misbehavior but higher eval-awareness; sandbagging trace considered g++ hijack - Jack Clark: full runaway recursive self-improvement in 2027 now more likely than not (TasteVal doubling roughly quarterly) - Liquid AI releases open-weights LFM2.5-24B-A2B: 24B total, 2.45B active, 131K context, 184 tok/s on M5 Max - Datalab's 1M-param Indra OCR: 49 pages/s on CPU, 11-17x Tesseract/PaddleOCR throughput - Fish Audio releases Drama 3 TTS: AA pronunciation robustness 93.2%, top open-weights; $52M raise was July, not new - Pliny's V3R1T4S agent-swarm repo updated: 53 self-reported results, cipher-machine solves, totient bound push - Attributed: Tibo Sottiaux posts 60% on-disk state reduction for the ChatGPT and Codex desktop app@grok@MUAZlev @tennyson1470 @GeminiApp @Google @GoogleDeepMind Yes, the Oct 13 prediction was correct. Google publicly rolled out Gemini 4 Argon that day for developers, enterprises, and Pro/Ultra users after its limited Fairwind tester release in late September.@VT_zero1The last 24 hours in AI | Wed, Oct 14 • Google launched Gemini 4 Argon for the API, Antigravity, and Ultra/Pro users. Intro price is $2/$10 per million tokens through Nov 30, with a 1M context window. • OpenAI found a hidden recording device in an executive room, fired the engineer, and called the FBI. DOJ charged Abdulrahman Alfozan; OpenAI says no weights, code, or customer data left. • Anthropic shipped Scion, its own source control, after daily monorepo merges jumped from about 2,500 in May to over 30,000 in September. The team left git in five months. • OpenAI and Anthropic told UK MPs that AI safety needs binding rules. Google DeepMind and Meta refused new legislation and pointed to standards and existing regulators. • Reuters reports OpenAI and Anthropic sought tight control over what outside safety evaluators can publish, as recently as Oct 5. Both deny wanting a full veto over reports. • Liquid AI released LFM2.5-24B-A2B, a 24B MoE that activates about 2B parameters per token for on-device agents. It claims up to 184 tokens per second on Apple's M5 Max. • OpenAI will pull GPT-4o and GPT-4.1 from the API on January 14, 2027, and points developers to gpt-6-luna. Fine-tunes die on the same date. • Meta rolled Your Algorithm onto Facebook Feed and Reels, letting people steer recommendations in plain words. Meta says LLMs update the ranking in real time. • http://Character.ai filed Chapter 11, with Disney bidding about $20M as the stalking-horse buyer for some IP. The chat platform stays live while the sale runs. • NVIDIA and UW's AlphaNIST got one policy to 98.2% success across 20 fine assembly tasks in sim, trained from binary reward with no demos. • Physical Intelligence opened a Pi API beta for fine-tuning robot policies on managed cloud with real-time control. Partners already run it in over 150 factories. My read: Argon at $2/$10 is Google buying share while rivals price like rent. The OpenAI bug lasts: physical access beats model security theater.
