News linked to both this project and an event.
According to TechFlow Research, Goldman Sachs' July 18 report pointed out that Moonshot AI released the Kimi K3 model, with 2.8 trillion parameters, surpassing Claude Fable 5 and GPT-5.6 Sol to top the Arena.ai coding leaderboard, with API pricing at $2.3 per million tokens setting a new high for Chinese models. Two days after the release, Zhipu AI fell 28%, MiniMax fell 16%, Nasdaq 100 index futures fell over 1.8%, and the Philadelphia Semiconductor Index cumulatively fell over 18% from highs. Goldman Sachs believes Kimi K3 marks a turning point: a Chinese lab unable to match the largest pre-training compute capacity in the West rapidly narrowed the gap with top US models through architectural innovation and reinforcement learning, proving that "scaling" is no longer the only winning path. Goldman Sachs warns that the "compute expansion era" may be ending, and the AI infrastructure investment logic built around "the more compute, the better" needs to be rewritten. Goldman Sachs maintains a Buy rating on MiniMax and Neutral on Zhipu AI. Future focus should be on the intensive launch of 2-5 trillion parameter models such as Zhipu GLM, Alibaba Qwen, and MiniMax M3 Pro.
analyst KawzInvests stated that Moonshot AI's upcoming Kimi K3 could become a significant event in the open-source AI space, and the infrastructure demand behind it may drive growth for AI cloud service platforms. Kimi K3 has approximately 2.8 trillion parameters, making it one of the ultra-large-scale open-source models. According to Moonshot's official evaluation, the model's performance is only slightly behind frontier models like Claude Fable 5 and GPT 5.6 Sol, and the full model weights are expected to be released on July 27.KawzInvests pointed out that a model of this scale cannot run on an ordinary laptop or even a single server; users need a computing cluster composed of a large number of GPUs to complete model loading and inference. When top-tier open-source models are made available for free, the biggest beneficiaries might not be ordinary users, but rather platforms that offer model hosting and inference services. For example, $DOCN (DigitalOcean) already supports serverless inference services for models like Kimi K2.6. Developers do not need to deploy hardware; they can call the model via API and pay per Token. Additionally, the platform hosts over 70 models and covers GPU leasing, model fine-tuning, and AI Agent development tools.As more large-scale open-source models are released, developers' demand for low-barrier AI infrastructure will continue to increase. Model hosting, inference services, and GPU cloud platforms may become key beneficiaries in the open-source AI wave.
OpenClaw founder Peter Steinberger posted on X that OpenClaw has officially released version v2026.7.1. This update was completed jointly by 532 contributors, with a cumulative total of 3,063 commits. Major updates include: comprehensive restructuring of the Web UI and new user onboarding process, significant optimization of iOS / Android / macOS clients, improved integration capabilities for Telegram, Slack, Discord, and iMessage, and added support for models such as GPT-5.6 and Muse Spark 1.1. In addition, this version simultaneously upgraded the Node version to fix automatic updater issues. If the automatic update fails, the upgrade can be completed manually via the Web installer.
B.AI 宣布 GPT-5.6 全系模型正式登陆 Web Chat，此前，B.AI API 已率先接入 GPT-5.6 Sol、GPT-5.6 Terra、GPT-5.6 Luna 三款模型，即日起用户无需配置 API，登录 https://chat.b.ai/chat 即可直接体验包括聊天问答、内容创作、代码编写与复杂推理在内的完整能力。此次全面开放标志着新一代模型在 B.AI 生态中的全场景落地，为开发者和普通用户提供“开箱即用”的智能化服务。未来，B.AI 将持续降低前沿 AI 技术的使用门槛，助力高效创作与决策。
According to Decrypt, OpenAI has released a new prompt engineering guide for the flagship model GPT-5.6 Sol, with the core concept being "Outcome-First Prompting"—clarifying goals and termination conditions and avoiding lengthy procedural instructions. Internal programming agent tests show that streamlined system prompts improved evaluation scores by approximately 10–15%, while reducing total token usage by 41–66% and costs by 33–67%. The guide also adds the text.verbosity API parameter to globally control output verbosity, as well as a new "Programmatic Tool Calls" section, supporting the handoff of intermediate tasks such as filtering and batch processing to code to reduce the model's judgment burden. Additionally, the guide warns that GPT-5.6 strictly follows prompt rules, and conflicting instructions within the prompt will consume significant inference tokens and may lead to erroneous output.
Haseeb posted on X, stating that with models like GLM 5.2, Fable, and GPT 5.6 already launched and actively used by attackers, DeFi has not experienced the anticipated "hacker apocalypse." Chart data shows that based on the current year's data and running rate, the annualized amount stolen from DeFi in 2026 is approximately $1.89 billion. The cumulative stolen amount for the year is around $986 million, lower than the 2025 level and still within the historical range. Haseeb noted that the deeper change now is that while the number of hacker attacks has increased, the scale of individual attacks is declining more rapidly. Attackers are increasingly targeting smaller protocols and abandoned projects, while large protocols have implemented more security enhancements. As a result, overall fund security has not significantly deteriorated.
Tibo, a core member of OpenAI Codex, posted on X platform stating that previously, it was too easy for users to enable high-compute mode without fully explaining its impact on usage quotas. The large-scale revamp of the desktop application made common features like chats and projects difficult to find. Additionally, promotional efforts overly focused on ChatGPT Work, leading some Codex users to mistakenly believe that Codex would be phased out in the future. Furthermore, the new version also caused issues with some multi-agent workflows, plugin functions, and other user experiences.In response to the above issues, OpenAI has already rolled out the first batch of improvement measures, including adjusting default settings and the model selector to prevent users from accidentally using high-cost configurations, fixing several plugin submission issues, optimizing the presentation of Codex within the product, and resolving some desktop experience problems.Additionally, over the past 24 hours, OpenAI has been continuously collecting user feedback, analyzing usage patterns, and communicating with users. Although the market has shown strong interest in GPT 5.6 Sol and the web and mobile versions of ChatGPT Work, OpenAI also acknowledges that this update has multiple issues.As of now, two rounds of Codex and ChatGPT Work usage quota top-ups have been completed. The next larger-scale update will be released next week, when chat and project features will return to the sidebar in a more familiar and customizable form. The platform will also enhance the visibility of usage quotas and reset times, further clarifying the respective applicable scenarios for ChatGPT Work and Codex.
B.AI announces that OpenAI's latest generation GPT-5.6 full lineup of models has officially been integrated into the B.AI API network. Effective immediately, developers can directly access three distinctly positioned models through B.AI's single interface: GPT-5.6 Sol specializes in ultimate reasoning and complex tasks; GPT-5.6 Terra is perfectly suited for daily workloads; GPT-5.6 Luna prioritizes ultra-fast response and cost-efficiency. Whether you are building agents, optimizing workflows, or exploring the boundaries of AGI applications, B.AI provides you with flexible and stable compute support. Log in to the B.AI platform now to unlock the full capabilities of GPT-5.6 and let the next generation of AI work for you.
OpenAI 表示，GPT-5.6 SOL、TERRA 和 LUNA 将于本周四公开发布，在全球范围内扩大预览访问权限。
the US Department of Commerce has approved OpenAI for the broad rollout of the GPT-5.6 model. According to insiders, OpenAI is expected to officially launch GPT-5.6 on a large scale this week, following the completion of additional safety tests and multiple rounds of meetings with US government officials. Previously, OpenAI was only permitted to release GPT-5.6 in phases to government-approved institutions, similar to the release restrictions previously imposed on Anthropic's Mythos and Fable models. (Axios)
According to Reuters citing The Information, SpaceXAI and Cursor plan to release their first jointly developed artificial intelligence model as early as Wednesday. The model was originally scheduled for launch early this week but was delayed for efficiency optimization. The report noted that the new model is expected to possess strong rapid information processing capabilities, with some performance aspects potentially competitive with Anthropic's Opus 4.8 and OpenAI's GPT 5.5.
Tech blogger Leo has revealed that OpenAI may make GPT-5.6 available to the public between July 7 and July 9, with the earliest possible date being July 7. The new model's plan usage quotas are said to be more generous, and OpenAI is further strengthening its safety strategies ahead of the launch.Additionally, according to sources, Google DeepMind has tentatively scheduled the release of Gemini 3.5 Pro for July 17. Another tech blogger, Astro Polo, stated that Gemini 3.5 Pro will support a 2 million token context window, doubling the 1 million token context window currently supported by Claude Sonnet 5, Claude Opus 4.8, and Claude Fable 5. This makes it more suitable for handling large codebases, lengthy documents, and long conversations.Note: The above information is based on market rumors and has not yet been officially confirmed by OpenAI or Google DeepMind.
PPP Prediction Market Tool monitoring shows that Polymarket's probability of "GPT-5.6 to be released to the public on July 7" has risen to 64%, up 26% in 24 hours.The rules state that a model explicitly named "GPT-5.6," or a model officially recognized as the direct successor to GPT-5.5 (such as GPT-5.7, GPT-5.8, etc.), will meet the settlement conditions. Task-specific models like GPT-Codex and Transcribe, cost-optimized versions such as Nano and Mini, and the o-series reasoning models are all included in this market. However, GPT-6 or other next-generation flagship models are not counted.Additionally, qualified models must be open to the public, including public beta or open waitlists; closed testing or private access only will not meet the requirements. The final settlement will primarily be based on OpenAI's official announcements and public information on its official website, and will be verified with mainstream media reports.On June 27, OpenAI launched the next-generation GPT-5.6 series of models, currently offering a limited preview to a select group of partners.Odaily Seer Channel continues to follow the prediction market, seeing changes before pricing.
According to Meituan's official release, Meituan has officially launched the new generation large model LongCat-2.0 and open-sourced it simultaneously. The model features a total of 1.6T parameters, making it the industry's first trillion-parameter model to complete full-process training and inference on a domestic computing cluster of 50,000 cards. It natively supports 1M ultra-long context and focuses primarily on code understanding, generation, and execution in Agentic Coding scenarios. Technically, LongCat-2.0 adopts the LongCat Sparse Attention (LSA) sparse attention mechanism, reducing long text computation complexity from quadratic to linear; achieves token-level dynamic activation (33B~56B) via a zero-computation expert mechanism; and introduces the MOPD architecture to fuse three sets of expert capabilities: Agent, Reasoning, and Interaction. In terms of training efficiency, the team spent three years overcoming challenges in adapting to domestic computing power, reducing the monthly average daily failure rate by over 70%, increasing training MFU by 1.5 times, and achieving steady-state daily throughput exceeding 1T tokens/day. In terms of performance evaluation, LongCat-2.0 achieved a score of 59.5 on SWE-bench Pro, surpassing Gemini 3.1 Pro (54.2), GPT-5.5 (58.6), and Claude Opus 4.6 (57.3); and achieved a score of 79.9 on BrowseComp.
Odaily OpenAI has officially launched the preview version of its next-generation GPT-5.6 series models, including the flagship model Sol, the balanced model Terra, and the fast, low-cost model Luna. GPT-5.6 introduces a new maximum reasoning effort and features a super mode that can accelerate complex tasks through sub-agents. The model boasts enhanced agentic capabilities in areas such as coding, biology, and cybersecurity. In terms of pricing, per 1 million tokens, Sol costs $5 for input and $30 for output; Terra costs $2.5 for input and $15 for output; Luna costs $1 for input and $6 for output.The model is currently in limited preview for a select group of trusted partners, with plans for a full rollout in the coming weeks. Additionally, the model will be available on Cerebras in July.
OpenAI has released its new GPT-5.6 series models—Sol, Terra, and Luna—currently available for limited preview to select partners. As the model names overlap with several well-known crypto projects, this has sparked market discussion.
after receiving a phone call from the Trump administration, OpenAI decided to restrict the deployment of GPT 5.6. In an internal memo, OpenAI CEO Sam Altman stated that the company would provide a limited preview version of GPT-5.6 to a small number of partners, with the U.S. federal government "approving customer access on a case-by-case basis." If the testing goes smoothly, the model will be publicly released on a broader scale in a few weeks. (TheGuardian)
Tech blogger @synthwavedd has stated that several highly anticipated cutting-edge AI models have adjusted their release schedules. Among them, GPT-5.6, originally slated for a near-term release, has been postponed to mid-July, while Google DeepMind has canceled the planned late-June launch of Gemini 3.5 Pro due to dissatisfaction with the model’s current performance.Meanwhile, preparations for the launch of OpenAI’s new-generation bidirectional voice model, Bidi, are progressing on the ChatGPT platform, with availability to users potentially coming as early as this week. Reports indicate that Bidi supports full-duplex voice interaction, allowing users to speak simultaneously with the model and interrupt the conversation at any time, marking a significant upgrade over existing voice modes.Additionally, Anthropic has granted early access to its Claude Sonnet 5 model for select enterprise clients. As the release progress of flagship models Mythos 5 and Fable 5 has slowed, Claude Sonnet 5 is seen as an interim product solution for Anthropic to alleviate competitive pressure.
OpenAI 推出 GPT-5.5-Cyber 网络安全模型，在 CyberGym 基准测试中得分为 85.6%，超越被美国政府下线的 Anthropic Mythos 5。
OpenAI announced an upgrade to its Daybreak network defense toolchain, officially launching the full version of the GPT-5.5-Cyber model tailored for cybersecurity defense scenarios.In the CyberGym benchmark test, which evaluates the ability of AI agents to reproduce known vulnerabilities, GPT-5.5-Cyber achieved a single-model score of 85.6%, surpassing GPT-5.5’s 81.8% and Anthropic Mythos 5’s 83.8%.OpenAI stated that as AI significantly enhances the efficiency of vulnerability discovery, the core bottleneck in cybersecurity is shifting from "finding vulnerabilities" to "automatically fixing them." To this end, the company has simultaneously upgraded the Codex Security plugin, enabling developers to automatically analyze vulnerabilities and generate fix patches within the Codex environment. Since the preview release in March this year, Codex Security has scanned over 30 million code commits and autonomously confirmed fixes for 500,000 security defects. The new version also supports CodeQL query integration and SARIF standard file export.Additionally, OpenAI has partnered with organizations such as Trail of Bits and HackerOne to launch the "Patch the Planet" open-source security project. This initiative provides ChatGPT Pro subscriptions and API credits to over 30 major open-source projects, including cURL and Go, with security expert teams manually verifying patch quality. Regarding the Daybreak partner program, security vendors such as Palo Alto Networks and Wiz have already integrated the relevant capabilities.