All previous Imagen models will shut down on June 24, 2026, with Google recommending migration to new Gemini Image models, including `gemini-2.5-flash-image`, `gemini-3.1-flash-image-preview`, and `gemini-3-pro-image-preview`.
Open source ↗SIGNAL TIMELINE
Signal Timeline
2026-06-24
As of June 19, 2026, the Gemini API will begin rejecting requests from unrestricted standard keys, requiring developers to use restricted standard API keys or migrate to more secure authorization keys; a full migration is mandated by September 2026.
Open source ↗As of June 24, 2026, OpenAI was reporting a "Minor OpenAI outage" impacting FedRAMP workspaces and API organizations with degraded performance.
Open source ↗@AndrewCurran_ mentioned xAI has a 10T Grok in training and that very large, high-performance models (including from xAI) are expected throughout the rest of 2026.
Open source ↗DeepMind is reportedly dissatisfied with the current state of Gemini 3.5 Pro, leading to its planned June 2026 launch being canceled; major releases are now expected in July.
Open source ↗@scaling01 followed up by noting the model ID "gpt-5-6-auto" in the context of the same leak, reinforcing speculation around near-term GPT-5.6 availability.
Open source ↗xAI has added official support for the Firecrawl plugin to Grok agents, enabling web scraping capabilities.
Open source ↗Speculation has emerged in the developer community that a "DeepSeek V4.1" update may be undergoing testing, following observations that the platform's fast mode knowledge cutoff was recently updated to June 2026.
Open source ↗Community users reported sudden and substantial performance improvements in the DeepSeek V4 Flash model, fueling rumors of an imminent new model release or refinement.
Open source ↗›View 12 more low-trust sources
As of June 24, 2026, DeepSeek's services were reported to be operational, with one user-submitted outage report in the preceding 24 hours.
Open source ↗Gemini 3.5 Pro is anticipated to launch in June 2026, following Google CEO Sundar Pichai's hint at a "next month" release during the Google I/O keynote.
Open source ↗Reports and leaks suggest that OpenAI's next immediate release will likely be an iterative update called GPT-5.6, rather than GPT-6.
Open source ↗A routing entry for 'gpt-5.6' was briefly observed in OpenAI's Codex backend logs in early June 2026, indicating active testing.
Open source ↗Internal codenames like 'kindle-alpha' have been associated with the development of OpenAI's next model.
Open source ↗OpenAI's chief scientist, Jakub Pachocki, reportedly described the company's next model as a 'meaningful improvement' over GPT-5.5, widely interpreted as referring to GPT-5.6.
Open source ↗Speculative timelines suggest a preview or announcement for GPT-6 could occur in Q4 2026, with general availability projected for Q1 2027.
Open source ↗There are hints of a potential late-June 2026 launch for GPT-5.6.
Open source ↗Gemini Live Translate was mentioned as one of the AI model releases in June 2026.
Open source ↗Multiple U.S. federal agencies and several states have restricted or banned the DeepSeek application on government devices and networks due to data-handling and national security concerns.
Open source ↗Microsoft is reportedly considering an Azure-hosted DeepSeek model as a lower-cost option for its Copilot Cowork enterprise agent.
Open source ↗DeepSeek has closed its first-ever external funding round, raising approximately 50 billion yuan (around $7.4 billion), which could significantly impact its development capabilities and market positioning.
Open source ↗2026-06-23
As of June 23, 2026, OpenAI's official status page indicated ongoing degraded performance for FedRAMP workspaces and API organizations, an issue that had persisted for about one week.
Open source ↗Qwen staff confirmed they will release open-weight versions of the Qwen 3.7 family of models.
Open source ↗According to AndrewCurran_, rumors persist about a new Sonnet model, and Anthropic teams are meeting frequently, with recent discussions possibly having gone well.
Open source ↗@scaling01 shared a screenshot from the ChatGPT website showing the model ID "gpt-5-6-auto" and commented "Did I just sleep through GPT-5.6 launch?", "looks like gpt-5.6 won't be delayed after all", and "2 leaks in a day is not a coincidence," suggesting an imminent or recent internal rollout of the next frontier model.
Open source ↗@scaling01 shared a screenshot from the ChatGPT website showing the model ID "gpt-5-6-auto" and commented "Did I just sleep through GPT-5.6 launch?", "looks like gpt-5.6 won't be delayed after all", and "2 leaks in a day is not a coincidence," suggesting an imminent or recent internal rollout of the next frontier model.
Open source ↗On the same day, @scaling01 posted a screenshot from the Codex GitHub repo spotting "GPT-5.6" and later noted "this suggests to me that GPT-5.6 could be an even more expensive/larger model than GPT-5.5," providing additional evidence of active development and testing of the next model.
Open source ↗On the same day, @scaling01 posted a screenshot from the Codex GitHub repo spotting "GPT-5.6" and later noted "this suggests to me that GPT-5.6 could be an even more expensive/larger model than GPT-5.5," providing additional evidence of active development and testing of the next model.
Open source ↗›View 8 more low-trust sources
A new Qwen model, 'Qwen/Qwen-AgentWorld-35B-A3B', was published on Hugging Face on June 23, 2026.
Open source ↗Effective June 23, 2026, Qoder introduced 'Off-Peak Rates' for Qwen3.7-Max and Qwen3.7-Plus, offering up to 80% savings during specific UTC hours daily for users.
Open source ↗Japan's Finance Minister confirmed on June 23, 2026, that the Japanese government and major domestic banks had gained access to Anthropic's Claude Mythos, but access was suspended following the U.S. government order.
Open source ↗A report from June 23, 2026, discussed a potential AI bubble in China due to computing power shortages, suggesting that Chinese AI companies, including Kimi, are "trying to stop you from buying" their models due to cost issues, indicating broader supply-side pressures.
Open source ↗Over the past week (before June 23, 2026), some Gemini CLI users experienced account disruptions due to Antigravity bans. Google is currently conducting a system-wide automated unban and introducing a new self-service reinstatement process.
Open source ↗As of June 23, 2026, at 10:18 AM PDT, Google AI Studio and Gemini API were reported as operational, despite some user-submitted reports of outages in the preceding 24 hours.
Open source ↗Kimi experienced user-reported service disruptions, with seven reports in the 24 hours leading up to June 23, 2026, though the overall system status remained operational.
Open source ↗As of June 23-24, 2026, xAI's official status pages indicate all Grok services (iOS, Android, Web, API) are fully operational without declared incidents, despite user reports of issues.
Open source ↗2026-06-22
GPT-5.6 / Bidi-1 is expected "this Thursday" (June 25), with extensive testing of 5.6-Pro already done, and other 5.6 checkpoints also tested.
Open source ↗According to AiBattle_, Claude Sonnet 5 is available to select enterprise customers via an Early Access Program as a stop-gap while the next major model's progress has stalled.
Open source ↗›View 5 more low-trust sources
StatusGator reported user-submitted outages for Z.ai's services on June 22, 2026, including error messages such as "Internal error: API Error: 529[The service may be temporarily overloaded, please try again later]" and prompts to "switch to GLM-5-Turbo for experience or try again later."
Open source ↗Google launched Gemini 2.5 Pro with Deep Think on June 22, 2026, described as its most capable model to date, featuring a 2 million token context window and an extended reasoning mode.
Open source ↗On June 22, 2026, OpenAI experienced an "Increased Error Rate for gpt-5.5 model in the API" and "Elevated transcription failures affecting ChatGPT & Codex," which were subsequently resolved.
Open source ↗Qwen's Qwen3.7-Plus has experienced user-reported service outages and disruptions in the past 24 hours as of June 22, 2026, including connection issues, internal errors, and extremely slow response times in various regions.
Open source ↗On June 22, 2026, monitoring service StatusGator reported the Grok xAI Website was up but noted 4 user-submitted outage reports in the past 24 hours, primarily from India, Indonesia, and the United Kingdom, which xAI did not officially declare as incidents.
Open source ↗2026-06-21
@chetaslua posted a comparison video of GPT-5.6 Pro vs. Fable 5, treating GPT-5.6 as an active upcoming model under testing.
Open source ↗It was speculated that Gemini 3.5 Pro could be released "next Thursday" (around June 26) alongside other model releases.
Open source ↗›View 2 more low-trust sources
Around June 21, 2026, Cursor announced a new 1.5-trillion-parameter model trained from scratch on over 100,000 Nvidia GPUs, with capabilities extending beyond coding.
Open source ↗DeepSeek experienced a minor outage on June 21, 2026, which temporarily disrupted user and API access for approximately 43 minutes before being resolved.
Open source ↗2026-06-20
JPMorgan expects Zhipu AI to launch its next flagship model, GLM-5.5, in August 2026, potentially as a >1T-parameter model.
Open source ↗›View 1 more low-trust sources
OpenAI is reportedly restricting access to its services for users in China, Russia, and Iran, with services slated to be cut off as of July 9, 2026.
Open source ↗2026-06-19
›View 3 more low-trust sources
On June 19, 2026, Google AI Studio and Gemini API experienced issues with Nano Banana 2 and Nano Banana Pro models when processing 2k or 4k resolution images, which have since been resolved.
Open source ↗Cursor co-founder Michael announced that the company's first fully self-trained AI model is currently undergoing training and is expected to ship "within the next few weeks."
Open source ↗Google AI Studio and Gemini API experienced a "Warning for 2 Hours and 35 Minutes" on June 19, 2026. User reports from around June 20-22, 2026, indicated issues such as "No image generation," "Error message," "Slow performance," and "Constant error regardless of the prompt."
Open source ↗2026-06-18
It was reported that Gemini 3.5 Pro was expected to be released mid-June.
Open source ↗›View 5 more low-trust sources
An incident affecting Grok access was resolved on June 18, 2026, after lasting over two hours.
Open source ↗xAI released Grok 4.3 on Amazon Bedrock, making it generally available to AWS users and expanding its reach.
Open source ↗An Enterprise release note on June 18, 2026, updated the status of Gemini 3.1 Pro and 3 Flash in the EU region from "currently available" to "coming soon."
Open source ↗The "Record & Replay" feature for the Codex app on macOS is initially unavailable in the European Union, UK, and Switzerland.
Open source ↗On June 18, 2026, Grok experienced an outage lasting approximately 135 minutes, leading to service interruptions.
Open source ↗2026-06-17
Gemini Omni Pro was mentioned in the context of editing capabilities, implying ongoing development of a next frontier variant.
Open source ↗A new Gemini image model codenamed "instant-ramen" was spotted in the Arena leaderboard, described as fast and native.
Open source ↗›View 5 more low-trust sources
Grok 4.3 became generally available on Amazon Bedrock on June 17, 2026, offering a 1-million-token context window and configurable reasoning levels.
Open source ↗On June 17, 2026, the US Department of Justice intervened in a lawsuit against xAI concerning unpermitted gas turbines at its Colossus 2 data center, arguing that halting these operations would endanger American national, economic, and energy security because Grok models support "mission-critical operations" for military and security agencies.
Open source ↗The U.S. government has temporarily refrained from adding Chinese AI startup DeepSeek to a key trade blacklist (Entity List) to avoid escalating tensions with Beijing. As of mid-June 2026, DeepSeek currently avoids direct U.S. export controls that would restrict access to U.S.-origin goods, software, or technology.
Open source ↗The US Commerce Department has reportedly delayed placing DeepSeek and over 100 other Chinese companies on its export control entity list. This decision, reported on June 17-18, 2026, suggests an effort by the US government to avoid further escalating tensions with Beijing.
Open source ↗The U.S. government has reportedly held off on adding DeepSeek to its trade-restricting Entity List to prioritize stability and avoid escalating geopolitical tensions with China.
Open source ↗2026-06-16
OpenAI updated its data residency information on June 16, 2026, confirming inference residency for ChatGPT in Europe (EEA + Switzerland), the United States, and the United Arab Emirates.
Open source ↗Starting June 16, 2026, the Gemini 3.5 Flash feature management toggle is no longer available in the Global, US, and EU multi-regions.
Open source ↗›View 4 more low-trust sources
Moonshot AI's status page reported an increased error rate for Agentic models on June 16, 2026.
Open source ↗Cloudflare Workers AI reported degraded availability for Kimi K2.6 and K2.7 Code models, with minor impact starting June 16, 2026.
Open source ↗On June 16, 2026, Cloudflare Workers AI reported degraded availability for some models, including Kimi K2.6 and K2.7-Code, which was under investigation.
Open source ↗DeepSeek completed its first round of external financing on June 16, 2026, raising over 50 billion RMB and valuing the company at approximately 50 billion USD. Founder Liang Wenfeng personally invested 20 billion RMB, with other investors including Tencent, CATL, JD.com, Alibaba, NetEase, and the National Artificial Intelligence Industry Investment Fund.
Open source ↗2026-06-15
On June 15, 2026, Google announced the deprecation of Imagen 4 and Gemini 3 Image models (to be shut down on August 17, 2026) and Veo video generation models (to be shut down on June 30, 2026).
Open source ↗›View 2 more low-trust sources
Anthropic suspended access to its Fable 5 and Mythos 5 AI models for all users following a U.S. government directive due to national security concerns, effective June 15, 2026. An official stated on June 24, 2026, that a Mythos model had identified vulnerabilities in sensitive U.S. government computer systems.
Open source ↗xAI released Grok Build 0.1, a fast coding model, via the xAI API in public beta on June 15, 2026.
Open source ↗2026-06-13
›View 1 more low-trust sources
DeepSeek experienced a seven-hour outage on its services, described as the largest service disruption since its debut.
Open source ↗2026-06-12
›View 1 more low-trust sources
Anthropic's Claude Fable 5 and Mythos 5 models were suspended on June 12, 2026, due to a U.S. government export control directive.
Open source ↗2026-06-10
According to the Gemma 4 model card published on June 10, 2026, the Gemma 4 family includes new model sizes such as E2B, E4B, 26B A4B, and 31B, which are multimodal open models supporting text and image input, with some variants also supporting audio input.
Open source ↗Gemini Pro and Flash models experienced a significant global outage on June 10, 2026, with approximately 90% of prompts failing for about 7 hours due to error codes 1076 and 1099.
Open source ↗›View 3 more low-trust sources
Google's Gemini experienced a significant global outage on June 10, 2026, with users reporting "Something went wrong" errors (codes 1099 and 1076) across multiple platforms. The outage was attributed to a performance issue in Google's backend database affecting the retrieval of Gemini App tools and was largely resolved by June 11, 2026.
Open source ↗Google DeepMind announced DiffusionGemma, an experimental open model that uses text diffusion methods and achieves up to 4x faster inference compared to typical local LLMs when using a GPU.
Open source ↗DiffusionGemma, an open-weight model released on June 10, 2026, allows drafting text in chunks for up to four times faster generation, optimized for rapid drafts and code editing, though with lower writing quality.
Open source ↗2026-06-09
Discussions from June 9, 2026 (with recent comments in the last 7 days) indicate that free access to Qwen Code CLI has been removed, and usage now requires an Alibaba Cloud Coding Plan.
Open source ↗2026-06-05
›View 1 more low-trust sources
SoftBank CEO Masayoshi Son, following discussions with Sam Altman and OpenAI engineers, stated that OpenAI's upcoming model is being designed by another AI model.
Open source ↗2026-06-03
Users have reported performance issues with DeepSeek's services over the past few weeks, noting that the service has become "unreliable" and "worse and worse," with one specific instance mentioning DeepSeek V4 Pro looping during an agentic OAuth integration task.
Open source ↗›View 1 more low-trust sources
On June 3, 2026, the UK's CMA issued binding rules requiring Google to allow publishers to opt out of content usage in AI Overviews; Google plans to implement these changes globally.
Open source ↗2026-06-01
›View 2 more low-trust sources
Several older Gemini 2.0 models (gemini-2.0-flash, gemini-2.0-flash-001, gemini-2.0-flash-lite, gemini-2.0-flash-lite-001) were officially shut down on June 1, 2026.
Open source ↗Google announced the general availability of the "Interactions API" in June 2026, unifying the handling of Gemini models and agents through a single interface, supporting state management and background execution for more "agent-standard" AI use cases.
Open source ↗2026-05-31
›View 2 more low-trust sources
Starting June 19, 2026, the Gemini API will gradually stop accepting requests from unrestricted API keys; users are advised to restrict their keys immediately.
Open source ↗Imagen 4 and Gemini 3 Image models are scheduled for deprecation and will be shut down on August 17, 2026.
Open source ↗2026-05-29
›View 1 more low-trust sources
Preview versions of Gemini 3.1 Flash Image and Gemini 3 Pro Image models are set for shutdown on June 25, 2026.
Open source ↗2026-05-21
›View 1 more low-trust sources
As of May 2026, Gemini implemented a new usage limit system with a "5-hour refresh" and "weekly limit," leading to increased reports of users reaching their limits and being automatically switched to lighter models.
Open source ↗2026-05-19
›View 3 more low-trust sources
Gemini Spark, a proactive personal AI agent, began rolling out to trusted testers on May 19, 2026, with a beta version for Google AI Ultra subscribers in the US expected in the coming weeks.
Open source ↗Google's Gemini 3.5 Pro is anticipated for general API availability in June 2026, with a full rollout expected between late June and early July. This was announced at Google I/O on May 19, 2026.
Open source ↗Google previewed Gemini 3.5 Pro at Google I/O 2026, with expected availability in June 2026, targeting advanced reasoning, complex coding, and research-intensive tasks.
Open source ↗2026-05-18
›View 1 more low-trust sources
Leaks prior to Google I/O 2026 mentioned Gemini 3.2 (Flash, Pro, Enterprise Ultra variants) appearing in Arena logs and API strings.
Open source ↗2026-05-14
Veo video generation models are being deprecated and will be shut down on June 30, 2026.
Open source ↗2024-06-23
According to chetaslua, while GPT-5.6's release has been delayed to mid-July, Bidi-1 preparations are continuing for a possible imminent release.
Open source ↗According to AiBattle_, the release of GPT-5.6 has been delayed until around mid-July.
Open source ↗2024-06-19
Grok 4.4 is expected to be released soon, with some users reporting it became available this week, though no screenshots have confirmed this.
Open source ↗2024-05-29
›View 1 more low-trust sources
DeepSeek secured a 50 billion yuan (approx. €6.4 billion) fundraising round, which founder Liang Wenfeng stated will finance the company's long-term research efforts for future models.
Open source ↗