{"schema_version":"1.0","slug":"agentic-ai-timing-2026","title":"Agentic AI 2026: Why Every Lab Moved at Once","description":"Agentic AI 2026: in four weeks Google, OpenAI and Anthropic shipped agent models at $2/$10, cut cache prices, moved into AWS, and OpenAI cut Pro allowances.","data_updated":"2026-10-05","source_post":"https://www.jsonhouse.com/posts/agentic-ai-timing-2026/","category":"industry-analysis","cluster":"CLUSTER_LLM","format":"A","attribution":{"source":"Json House","source_url":"https://www.jsonhouse.com/posts/agentic-ai-timing-2026/","dataset_url":"https://www.jsonhouse.com/data/agentic-ai-timing-2026.json","citation":"Json House, \"Agentic AI 2026: Why Every Lab Moved at Once\", jsonhouse.com (2026-10-05)","attribution_required":true,"terms_url":"https://www.jsonhouse.com/data-policy/"},"key_facts":[{"fact":"Gemini 4 Argon, announced 2026-09-30, raises the output limit from 64,000 to 1 million tokens and is priced at an introductory $2 input / $10 output per 1M tokens with cached input 95% off, rising to $4 / $20 afterwards; the end date of the introductory rate is not published","source":"Google, Gemini 4 Argon announcement, retrieved 2026-10-05","category":"data"},{"fact":"Four tracked models sit at exactly $2 input / $10 output per 1M tokens on 2026-10-05 — Claude Sonnet 5, Claude Sonnet 5.5, GPT-6-sol and GPT-6.1-sol — against one (Claude Sonnet 5) on 2026-07-16","source":"jsonhouse weekly pricing snapshot series, 2026-07-16 and 2026-10-05","category":"data"},{"fact":"Until 2026-09-01 the only tracked models pricing a cache read below 0.1x of their own input price were DeepSeek's; by 2026-10-05 four US models had followed (Claude Fable 5.1, Claude Mythos 5.1, Claude Opus 5.5, GPT-6.1-sol)","source":"jsonhouse weekly pricing snapshot series, 2026-07-16 through 2026-10-05","category":"trend"},{"fact":"OpenAI's cheapest sol model cost 11.5x DeepSeek v4-pro's off-peak input price on 2026-07-16 and 3.0x on 2026-10-05; the gap closed from both sides, with the sol input price down 60% and DeepSeek v4-pro up 52%","source":"jsonhouse derived from weekly pricing snapshots of OpenAI and DeepSeek official pricing pages","category":"data"},{"fact":"Anthropic measured AI agents using about 4x the tokens of chat interactions and multi-agent systems about 15x, and found token usage alone explained 80% of performance variance in its browsing evaluation","source":"Anthropic Engineering, multi-agent research system, June 2025","category":"data"},{"fact":"OpenAI reopened ChatGPT Pro 200 on 2026-09-29 at an unchanged $200 per month with a lower usage allowance for new subscriptions; eligible existing subscribers keep the previous allowance only through 2026-10-29. OpenAI's pages do not state the size of the reduction","source":"OpenAI Help Center, About ChatGPT Pro tiers, retrieved 2026-10-05","category":"policy"},{"fact":"GPT-6 Astra became generally available on Amazon Bedrock on 2026-09-08 and GPT-6 Sol and Luna on 2026-09-22; Claude Platform became available inside AWS accounts on 2026-05-11","source":"AWS What's New announcements","category":"trend"},{"fact":"Meta uses interactions with Meta AI to personalize content and ads from 2025-12-16, excluding topics such as religious views, sexual orientation, political views and health from ad targeting; Meta states more than 1 billion people use Meta AI monthly","source":"Meta Newsroom, October 1, 2025","category":"policy"},{"fact":"Anthropic's consumer terms let Claude Free, Pro and Max users choose whether their chats train models, with 5-year retention if allowed and 30 days if not, and do not apply to API or commercial use including via Amazon Bedrock and Google Cloud Vertex AI","source":"Anthropic, Updates to Consumer Terms and Privacy Policy, August 28, 2025","category":"policy"}],"faq_summary":[{"q":"Why is agentic AI taking off in 2026?","a":"Three things changed in the same four weeks: models built for long-horizon work shipped, the cache-read price agents mostly pay fell below 10% of input at four US models, and US models became purchasable inside AWS accounts."},{"q":"What is Gemini 4 Argon and how much does it cost?","a":"Google's frontier model announced 2026-09-30 with a 1-million-token output limit. Introductory pricing is $2 / $10 per 1M tokens with cached input 95% off, then $4 / $20; the end date is not published."},{"q":"Did OpenAI raise the price of ChatGPT Pro?","a":"Not the price. Pro 200 stays at $200 a month, but new subscriptions not eligible for grandfathering get a lower allowance, and eligible subscribers keep the previous one only through 2026-10-29. OpenAI's pages do not state the size of the cut."},{"q":"Why does the cache price matter more than the list price for AI agents?","a":"Agents re-send instructions, tools and history on every step, using about 4x the tokens of a chat. On that volume the cache-read price decides the bill."},{"q":"Will AI agents use my personal data for ads or training?","a":"It depends on the tier. Meta uses Meta AI interactions for content and ad personalization; Anthropic lets consumers choose on training and exempts API and commercial use; enterprise agents on Amazon Bedrock are processed inside Bedrock."}],"primary_sources":[{"title":"Gemini 4 Argon announcement","url":"https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/","publisher":"Google"},{"title":"About ChatGPT Pro tiers","url":"https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers","publisher":"OpenAI"},{"title":"OpenAI models, Codex, and Managed Agents come to AWS","url":"https://openai.com/index/openai-on-aws/","publisher":"OpenAI"},{"title":"OpenAI GPT-6 Astra is now generally available on Amazon Bedrock","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/openai-gpt-6-astra-on-amazon-bedrock/","publisher":"AWS"},{"title":"OpenAI GPT-6 Sol and GPT-6 Luna are now generally available on Amazon Bedrock","url":"https://aws.amazon.com/about-aws/whats-new/2026/09/openai-gpt-6-sol-luna-on-amazon-bedrock/","publisher":"AWS"},{"title":"Claude Platform on AWS is now generally available","url":"https://aws.amazon.com/about-aws/whats-new/2026/05/claude-platform-aws/","publisher":"AWS"},{"title":"How we built our multi-agent research system","url":"https://www.anthropic.com/engineering/multi-agent-research-system","publisher":"Anthropic"},{"title":"Improving Your Recommendations on Our Apps With AI at Meta","url":"https://about.fb.com/news/2025/10/improving-your-recommendations-apps-ai-meta/","publisher":"Meta"},{"title":"Updates to Consumer Terms and Privacy Policy","url":"https://www.anthropic.com/news/updates-to-our-consumer-terms","publisher":"Anthropic"},{"title":"Claude API pricing","url":"https://platform.claude.com/docs/en/about-claude/pricing","publisher":"Anthropic"},{"title":"DeepSeek API pricing","url":"https://api-docs.deepseek.com/quick_start/pricing","publisher":"DeepSeek"}],"timeline":[{"date":"2026-09-07","event":"Claude Fable 5.1 first observed with cache read at 0.025x input (observed date, not launch date)","source":"Anthropic pricing page via jsonhouse weekly snapshot"},{"date":"2026-09-08","event":"GPT-6 Astra generally available on Amazon Bedrock","source":"AWS What's New"},{"date":"2026-09-14","event":"deepseek-flash first observed at $0.15 / $0.60 with cache read $0.003 and image input (observed date)","source":"DeepSeek pricing page via jsonhouse weekly snapshot"},{"date":"2026-09-22","event":"GPT-6 Sol and Luna generally available on Amazon Bedrock","source":"AWS What's New"},{"date":"2026-09-28","event":"Claude Opus 5.5 at $4 / $20 (0.05x cache) and GPT-6-sol at $2 / $10 first observed (observed date)","source":"Anthropic and OpenAI pricing pages via jsonhouse weekly snapshot"},{"date":"2026-09-29","event":"OpenAI reopens Pro 200 with a lower allowance for new subscriptions and launches Pro 500","source":"OpenAI Help Center"},{"date":"2026-09-30","event":"Google announces Gemini 4 Argon with a 1M-token output limit, $2 / $10 introductory pricing, Fairwind Program rollout","source":"Google"},{"date":"2026-10-05","event":"Claude Sonnet 5.5 at $2 / $10 and GPT-6.1-sol at $2 / $10 with cache read $0.10 first observed (observed date)","source":"Anthropic and OpenAI pricing pages via jsonhouse weekly snapshot"}],"comparison_data":{"field_provenance":{"input_per_1m_usd":"vendor","output_per_1m_usd":"vendor","cache_read_per_1m_usd":"vendor, except Gemini 4 Argon which is jsonhouse derived from Google's stated 95% discount","cache_discount_pct":"jsonhouse derived","status":"vendor statement or jsonhouse first-observed date"},"dimensions":["vendor","input_per_1m_usd","output_per_1m_usd","cache_read_per_1m_usd","cache_discount_pct","status"],"unit":"USD per 1M tokens, standard non-batch tier; tracked models collected 2026-10-05, Gemini 4 Argon from Google's 2026-09-30 announcement","entries":[{"model":"Claude Sonnet 5","vendor":"Anthropic","input_per_1m_usd":2.0,"output_per_1m_usd":10.0,"cache_read_per_1m_usd":0.2,"cache_discount_pct":90.0,"status":"Standard price since 2026-08-17"},{"model":"Claude Sonnet 5.5","vendor":"Anthropic","input_per_1m_usd":2.0,"output_per_1m_usd":10.0,"cache_read_per_1m_usd":0.2,"cache_discount_pct":90.0,"status":"First observed 2026-10-05"},{"model":"GPT-6-sol","vendor":"OpenAI","input_per_1m_usd":2.0,"output_per_1m_usd":10.0,"cache_read_per_1m_usd":0.2,"cache_discount_pct":90.0,"status":"First observed 2026-09-28"},{"model":"GPT-6.1-sol","vendor":"OpenAI","input_per_1m_usd":2.0,"output_per_1m_usd":10.0,"cache_read_per_1m_usd":0.1,"cache_discount_pct":95.0,"status":"First observed 2026-10-05"},{"model":"Gemini 4 Argon","vendor":"Google","input_per_1m_usd":2.0,"output_per_1m_usd":10.0,"cache_read_per_1m_usd":0.1,"cache_discount_pct":95.0,"status":"Introductory price; Fairwind Program only; regular $4 / $20, end date not published; not on Google's public pricing page as of 2026-10-05"}]},"numerical_data":{"field_provenance":"all metrics jsonhouse derived from vendor-published prices in the weekly snapshot series","metrics":[{"name":"models_below_0_1x_cache_read","values":{"2026-07-16":2,"2026-09-01":2,"2026-09-07":4,"2026-09-28":5,"2026-10-05":6},"unit":"tracked models","note":"US-vendor models among them: 0, 0, 2, 3, 4"},{"name":"input_ratio_cheapest_openai_sol_to_deepseek_v4_pro","values":{"2026-07-16":11.5,"2026-10-05":3.0},"unit":"x","note":"GPT-5.6-sol $5.00 vs $0.435 on 2026-07-16; GPT-6.1-sol $2.00 vs $0.66 on 2026-10-05; DeepSeek off-peak"},{"name":"output_ratio_cheapest_openai_sol_to_deepseek_v4_pro","values":{"2026-07-16":34.5,"2026-10-05":5.1},"unit":"x","note":"$30.00 vs $0.87; $10.00 vs $1.98"},{"name":"cache_read_ratio_cheapest_openai_sol_to_deepseek_v4_pro","values":{"2026-07-16":138,"2026-10-05":4.5},"unit":"x","note":"$0.50 vs $0.003625; $0.10 vs $0.022"},{"name":"models_at_2_input_10_output","values":{"2026-07-16":1,"2026-10-05":4},"unit":"tracked models","note":"Gemini 4 Argon's introductory rate is also $2 / $10 but is outside the tracked series"}]}}