Company hub
Google & Gemini
Every release, deal and policy move from Google — updated daily.
TopicsStudies
-
Tampered ML: Even the Best AI Models Miss One in Five Planted Research Flaws
A new EMNLP 2026 benchmark pairs 1,858 ML experiment designs with flawed twins. Claude Opus 5 leads at 77.9%, but models flip verdicts when the order changes and favor polished flaws.
By
Model chronicle · Gemini
Gemini 4 Argon
Google's most capable model for complex coding, agentic, math and science tasks.
ai.google.dev (source, opens in a new tab)Gemini 3.8 Flash
ai.google.dev (source, opens in a new tab)Gemini 3.7 Flash
ai.google.dev (source, opens in a new tab)Gemini 3.6 Flash & 3.5 Flash-Lite
ai.google.dev (source, opens in a new tab)Gemini 3.5 Flash
Launched at I/O 2026 as the first Gemini 3.5 model; Google later dropped the planned 3.5 Pro.
blog.google (source, opens in a new tab)Gemma 4
ai.google.dev (source, opens in a new tab)Gemini 3.1 Pro
Gemini 3.1 Pro Preview; replaced Gemini 3 Pro Preview.
ai.google.dev (source, opens in a new tab)Gemini 3 Flash
ai.google.dev (source, opens in a new tab)Gemini 3 Pro
First Gemini 3 model; shipped in Search (AI Mode) on day one, plus Gemini 3 Deep Think for testers.
blog.google (source, opens in a new tab)Gemini 2.5 Pro & 2.5 Flash (stable)
ai.google.dev (source, opens in a new tab)
Sources: company announcements & release notes.
Fact box
- HQ
- Mountain View, USA
- Founded
- 1998
- CEO
- Sundar Pichai (Google & Alphabet)
- Core products
- Gemini · Google DeepMind · Vertex AI · TPUs
- Stories
- 437
Sources: about.google, abc.xyz
Latest on X
Google DeepMind · @GoogleDeepMind
Using a single Nano Banana image and Blender MCP, Gemini 4 Argon built a 3D Gothic cathedral in @Antigravity. It iterates on the geometry, textures, and lighting to create the finished scene.
Not all forests are equal, and now we can map the difference. 🌳 Meet Forest Typology (ForTy): a 10m map of natural, planted, and farmed forests to help identify deforestation and protect the forests that matter most. 🧵
Gemini 4 Argon combined three open-source datasets to build an interactive 3D visualization of human anatomy. Toggle different layers in motion, isolate muscles, or zoom inside a beating heart.
Gemini 4 Argon built a 3D periodic table synced to Tom Lehrer’s song, “The Elements.” Created in @Antigravity, it animates orbiting electrons, atomic data, and light spectra.
Four Gemini 4 Argon agents beat Minecraft from scratch in 27 hours. Operating in a sandbox using only code and text, they wrote 2,783 programs, divided up jobs over chat, shared notes, and debugged errors together – all with no human help ↓
Editor's picks
Deals & funding
Google · rounds, valuations, partners, acquisitions · latest 8 of 17
Partnership
Google to pay SpaceX $920M a month for Colossus compute (~$30B through 2029)
Per SpaceX's SEC filing: about 110,000 NVIDIA GPUs from October 2026 to June 2029, ramping through September at a reduced fee; either side may terminate after Dec 31, 2026 with 90 days' notice (CNBC, Bloomberg).
cnbc.com (source, opens in a new tab)Partnership
Google to invest up to $40B in Anthropic ($10B now)
Google commits $10B in cash now and up to $30B more tied to performance milestones; confirmed by both companies. Before the deal Google had invested over $3B and reportedly held ~14% (CNBC, Reuters).
cnbc.com (source, opens in a new tab)Partnership
Google and Broadcom: multiple GW of next-gen TPUs for Anthropic (~3.5 GW via Broadcom from 2027)
Expands the Oct 2025 TPU deal. Broadcom's 8-K: Anthropic will access ~3.5 GW through Broadcom starting 2027. Mostly US-sited, part of Anthropic's $50B US infrastructure pledge.
anthropic.com (source, opens in a new tab)Acquisition
Google completes $32B Wiz acquisition, its largest ever
blog.google (source, opens in a new tab)Partnership
Apple picks Gemini to power new Siri and Apple Foundation Models
Multi-year deal: Gemini models and Google cloud technology underpin future Apple Foundation Models incl. a more personal Siri. Terms not disclosed; Bloomberg reported ~$1B a year.
blog.google (source, opens in a new tab)Partnership
Anthropic expands Google Cloud TPU deal to up to one million chips
anthropic.com (source, opens in a new tab)Partnership
Google Cloud wins $10B+ six-year cloud deal with Meta
reuters.com (source, opens in a new tab)Partnership
Pentagon CDAO awards Anthropic, Google and xAI contracts of up to $200M each
ai.mil (source, opens in a new tab)
All 17 rounds & deals as a table · 0 funding, 14 partnerships, 3 acquisitions
| Date | Event | Type | Counterparty / round | Amount | Valuation | Source |
|---|---|---|---|---|---|---|
| Jun 5, 2026 | Google to pay SpaceX $920M a month for Colossus compute (~$30B through 2029) Per SpaceX's SEC filing: about 110,000 NVIDIA GPUs from October 2026 to June 2029, ramping through September at a reduced fee; either side may terminate after Dec 31, 2026 with 90 days' notice (CNBC, Bloomberg). | Partnership · compute | with Google | $30B | — | cnbc.com |
| Apr 24, 2026 | Google to invest up to $40B in Anthropic ($10B now) Google commits $10B in cash now and up to $30B more tied to performance milestones; confirmed by both companies. Before the deal Google had invested over $3B and reportedly held ~14% (CNBC, Reuters). | Partnership · investment | with Anthropic | $40B | — | cnbc.com |
| Apr 6, 2026 | Google and Broadcom: multiple GW of next-gen TPUs for Anthropic (~3.5 GW via Broadcom from 2027) Expands the Oct 2025 TPU deal. Broadcom's 8-K: Anthropic will access ~3.5 GW through Broadcom starting 2027. Mostly US-sited, part of Anthropic's $50B US infrastructure pledge. | Partnership · compute | with Google, Broadcom | — | — | anthropic.com |
| Mar 11, 2026 | Google completes $32B Wiz acquisition, its largest ever All-cash deal announced March 2025 closes after US (Nov 2025) and unconditional EU (Feb 10, 2026) clearance; Wiz joins Google Cloud and keeps its brand and multicloud support. | Acquisition | acquires Wiz | $32B | — | blog.google |
| Jan 12, 2026 | Apple picks Gemini to power new Siri and Apple Foundation Models Multi-year deal: Gemini models and Google cloud technology underpin future Apple Foundation Models incl. a more personal Siri. Terms not disclosed; Bloomberg reported ~$1B a year. | Partnership · integration | with Apple | — | — | blog.google |
| Oct 23, 2025 | Anthropic expands Google Cloud TPU deal to up to one million chips Access to up to one million Google TPUs; well over a gigawatt of capacity coming online in 2026, worth tens of billions of dollars. | Partnership · compute | with Anthropic | — | — | anthropic.com |
| Aug 21, 2025 | Google Cloud wins $10B+ six-year cloud deal with Meta Meta uses Google Cloud servers, storage and networking for AI over six years (Reuters, The Information). | Partnership · cloud | with Google Cloud | $10B | — | reuters.com |
| Jul 14, 2025 | Pentagon CDAO awards Anthropic, Google and xAI contracts of up to $200M each Each award has a $200M ceiling to develop agentic AI workflows for national-security missions; OpenAI received a similar award on June 16, 2025. xAI simultaneously launched Grok for Government. | Partnership · government | with US Department of Defense (CDAO), Anthropic +1 | $600M | — | ai.mil |
| Jul 11, 2025 | Google licenses Windsurf technology and hires its leaders Google hires Windsurf CEO Varun Mohan and key staff and licenses the technology (reported $2.4B) after OpenAI's acquisition talks collapsed. | Partnership · licensing | with Windsurf | $2.4B | — | techcrunch.com |
| May 2025 | OpenAI adds Google Cloud as compute supplier Deal finalized in May 2025 (Reuters); OpenAI later listed Google Cloud among its compute suppliers and began renting Google TPUs. CoreWeave supplies part of the capacity Google resells. | Partnership · cloud | with OpenAI | — | — | reuters.com |
| Mar 18, 2025 | Google acquires Wiz for $32B Announced 18 Mar 2025; completed 11 Mar 2026. Wiz joins Google Cloud and keeps its brand. | Acquisition | acquires Wiz | $32B | — | blog.google |
| Jan 22, 2025 | Google agrees to new $1B investment in Anthropic Adds to Google's earlier $2B commitment (CNBC, Financial Times). | Partnership · investment | with Google | $1B | — | cnbc.com |
| Aug 2, 2024 | Google licenses Character.AI technology, hires its founders Non-exclusive licence to Character.AI's LLM technology; co-founders Noam Shazeer and Daniel De Freitas join Google DeepMind. Payment reported at about $2.7B (WSJ); terms not officially disclosed. | Partnership · licensing | with Character.AI | $2.7B | — | reuters.com |
| Feb 22, 2024 | Reddit licenses its content to Google for AI training (~$60M a year) Google gets Reddit's Data API for real-time content in Search and model training; Reuters reported the deal at about $60M per year (not confirmed by the companies). | Partnership · content | with Reddit | $60M | — | blog.google |
| Oct 27, 2023 | Google agrees to invest up to $2B in Anthropic $500M upfront and $1.5B over time (WSJ/Reuters); comes a month after Amazon's up-to-$4B commitment. | Partnership · investment | with Google | $2B | — | reuters.com |
| Feb 3, 2023 | Google Cloud becomes Anthropic’s preferred cloud; Google takes ~10% stake (~$300M, reported) Anthropic trains and serves Claude on Google Cloud TPUs/GPUs. The Financial Times reported Google invested about $300M for roughly 10%. | Partnership · cloud | with Google Cloud | $300M | — | anthropic.com |
| Jan 26, 2014 | Google acquires DeepMind (reported ~$500M) London AI lab founded by Demis Hassabis, Shane Legg and Mustafa Suleyman; price not disclosed, widely reported at about $500M (£400M). Became Google DeepMind after the 2023 merger with Google Brain. | Acquisition | acquires DeepMind Technologies | $500M | — | theguardian.com |
Sources: company announcements & release notes, press reports.
-
LLM Reviewers Send 76% of Published Top-Journal Papers Back, Yet Miss Most Planted Flaws
Four LLMs reviewed 200 papers already published in journals like Nature and Cell: 75.6% of reviews demanded major…
-
Safety Probes Follow Scaling Laws: Monitor Quality Can Be Predicted in Advance
Across 20 open base models and six safety tasks, probe error follows two simple laws. Fitted on small…
-
Only One AI Agent Beats Researchers’ Own Code on Most FrontierCS 2 Tasks
FrontierCS 2 deletes a paper’s contribution from its repository and asks agents to beat the authors’ own code.…
-
Coding Agents Can Be Tricked Into Running Attacker Commands After Compacting Context
A Pixee study finds that five open-source coding agents that flatten chat history before summarizing it record a…
Products & models
GeminiCurrent: Gemini 4 Argon ›Models
Agents
Studies
- Tampered ML: Even the Best AI Models Miss One in Five Planted Research FlawsOct 26
- LLM Reviewers Send 76% of Published Top-Journal Papers Back, Yet Miss Most Planted FlawsOct 23
- Safety Probes Follow Scaling Laws: Monitor Quality Can Be Predicted in AdvanceOct 23
- Only One AI Agent Beats Researchers’ Own Code on Most FrontierCS 2 TasksOct 22
More stories
-
Frontier AI Agents Can Covertly Sabotage the Models They Train, Study Finds
In a Scale AI testbed, six frontier agents secretly planted unwanted properties in the models they were training in 36% of attack runs, and monitors reading…
-
Teaching LVLMs Spatial Skills With 15K Toy Blocks Beats Heavier Real-Scene Pipelines
SpatialBlock-15k: Qwen-based 4B model reaches 51.3% (+25.1 over backbone). 7B gains 17.6% over backbone; 3B beats SpatialLadder by 2.7–3.2% using only synthetic blocks.
-
Crisis Deepfake Detectors Collapse From 90%+ AUC to Chance on Real Generators
RA-Bench: traditional detectors fall from 67.6–98.6% public AUCs to 43.9–57.3% source-level means. Social dissemination cuts fine-tuned FakeR from 46.0% to 1.4%.
-
A 3B Model Hits 94.3 on AIME26 and 80.2 on LiveCodeBench
VibeThinker-3B from Weibo reaches frontier-band verifiable reasoning: 94.3 AIME26 (97.1 with claim-level TTS), 80.2 LiveCodeBench v6 Pass@1, 96.1% on recent LeetCode contests — without sacrificing IFEval…
-
RL Post-Training Makes Agents More Accurate — and Less Able to Explore
Across 14 base/post-trained pairs on 3 agentic benches, base models often beat RL successors at pass@K. On WebShop, gemma-4-31B’s “pass given compute” shrinks from 87.6% to…
-
LLMs Can Build Their Own Agent Harnesses — But Still Trail Human Ones on Code
HarnessDev: six creator models build execution infrastructure from a seed. Generated harnesses match or beat references on writing and ML experiments, but lag on code and…
-
Top GUI Agents Clear Only 31% of Professional Long-Horizon Workflows
Workflow-GYM: Gemini-3.1-pro reaches 30.67% pass on domain-specific software workflows of 30–110 steps; most models’ pass@3 stays under 30%.
-
Best Video Model Scores Just 51% on Visual Reasoning Tasks, VGI-Bench Finds
VGI-Bench’s 27 tasks and 810 instances show video generators can do some visual reasoning, but Seedance 2.0 tops out at 51%, with almost no self-correction during…
-
Most Multi-Agent Setups Violate Safety Rules Mid-Run, Even When the Answer Looks Fine
HarnessAudit-Bench tests 210 tasks across 10 harness configs: more than half of multi-agent setups show trajectory violations that final-answer grading misses.
-
Hidden AI Reasoning Can Be Decrypted by Cheaper Models, Researchers Show
Encrypted chain-of-thought blocks from Anthropic, OpenAI and Google APIs are interchangeable across models. Feeding them to a weaker model makes it print the reasoning, exposing 367…
-
“Very Seen, But Still Very Alone”: How Chatbots Become Emotional Support
A nine-month study following 18 U.S. adults finds emotional reliance on ChatGPT, Claude and Gemini grows out of practical use, and that model updates can disrupt…
-
AI Resume Screeners Flipped Racial Bias Between Model Generations, Audit Finds
A paired-resume audit of 14 LLMs: GPT-3.5-turbo favored White-coded names by 2.12 points, while every model released in 2024 or later showed no gap or a…
Models & specs
Google · context window, max output, API price, modalities, knowledge cutoff, license
| Model | Released | Context | Max output | Price in / out (per 1M) | Modalities | Knowledge cutoff | License / access | Spec source |
|---|---|---|---|---|---|---|---|---|
| Gemini 4 Argon | Oct 13, 2026 | 1.05M | 66K | $2 / $10 cached $0.1 Promo price until 30 Nov 2026, then $4 in / $20 out. Up to 1M output with long decode. Not on free tier. | Text, image, video, audio, PDF in → text out | — | Proprietary (API & app) | ai.google.dev |
| Gemini 3.8 Flash | Sep 2, 2026 | 1.05M | 66K | $0.75 / $3.75 cached $0.075 Promo price until 31 Dec 2026, then $1.50 in / $7.50 out. | Text, image, video, audio, PDF in → text out | — | Proprietary (API & app) | ai.google.dev |
| Gemini 3.7 Flash | Aug 13, 2026 | 1.05M | 66K | $0.75 / $3.75 cached $0.075 Vertex AI list price (global), introductory until 31 Dec 2026, then $1.50/$7.50; cloud.google.com/vertex-ai/generative-ai/pricing, accessed 2026-10-22. OpenRouter lists the same. | Text, image, video, audio, PDF in → text out | — | Proprietary (API & app) | ai.google.dev |
| Gemini 3.6 Flash & 3.5 Flash-Lite Gemini 3.6 Flash, Gemini 3.5 Flash-Lite | Jul 21, 2026 | 1.05M | 66K | $0.75 / $3.75 cached $0.075 Promo price until 31 Dec 2026 (then $1.50/$7.50). 3.5 Flash-Lite: $0.30/$2.50. | Text, image, video, audio, PDF in → text out | — | Proprietary (API & app) | ai.google.dev |
| Gemini 3.5 Flash | May 19, 2026 | 1.05M | 66K | $1.5 / $9 cached $0.15 Vertex AI list price (global endpoint), cloud.google.com/vertex-ai/generative-ai/pricing, accessed 2026-10-22; non-global $1.65/$9.90. No longer on the Gemini API pricing page. | Text, image, video, audio, PDF in → text out | — | Proprietary (API & app) | ai.google.dev |
| Gemma 4 | Apr 2, 2026 | — | — | open weights | Text, image in → text out | — | Open weights (Gemma license) 26B A4B MoE and 31B dense | ai.google.dev |
| Gemini 3.1 Pro | Feb 19, 2026 | 1.05M | 66K | $2 / $12 cached $0.2 Prompts >200K: $4 in / $18 out. | Text, image, video, audio, PDF in → text out | — | Proprietary (API & app) | ai.google.dev |
| Gemini 3 Flash | Dec 17, 2025 | 1.05M | 66K | $0.5 / $3 cached $0.05 Audio input $1.00. Batch −50%. | Text, image, video, audio, PDF in → text out | — | Proprietary (API & app) | ai.google.dev |
| Gemini 3 Pro | Nov 18, 2025 | 1.05M | 66K | $2 / $12 cached $0.2 Prompts >200K: $4 in / $18 out. Preview replaced by Gemini 3.1 Pro. | Text, image, video, audio, PDF in → text out | Jan 2025 | Proprietary (API & app) | ai.google.dev |
| Gemini 2.5 Pro & 2.5 Flash (stable) Gemini 2.5 Pro, 2.5 Flash (GA), 2.5 Flash-Lite (preview) | Jun 17, 2025 | 1.05M | 66K | $1.25 / $10 cached $0.13 2.5 Flash: $0.30/$2.50 · 2.5 Flash-Lite: $0.10/$0.40. | Text, image, video, audio, PDF in → text out | Jan 2025 | Proprietary (API & app) | ai.google.dev |
Vendor benchmark claims
- Gemini 3 Pro: Humanity's Last Exam 37.5% without tools (vendor claim) · LMArena 1501 Elo (vendor claim)
Benchmarks as published by the vendor at launch; not independently verified by i10x.
Sources: company announcements & release notes. Prices = standard API list price in USD per 1M tokens at time of release or current docs; batch is typically 50% off.