Collected in Hong Kong time
AI Model Daily - 2026-05-25
2026-05-25 AI model news: no new flagship launch appeared in the last 24 hours; the main signals are Claude Mythos Preview capability validation, Grok Build agentic coding access, Qwen3.7-Max availability, and small video leaderboard changes.
30-Second Summary
- No new flagship model announcement from OpenAI, Google, Anthropic, Meta, Mistral, or Qwen appeared in the last 24 hours. Today's signals are mostly capability validation, distribution, and leaderboard drift.
- Anthropic
Claude Mythos Previewdeserves higher priority: the Project Glasswing initial update says Mythos Preview scanned 1,000+ open-source projects and estimated 23,019 vulnerabilities, including 6,202 high- or critical-severity items. This is not a new release, but it is a strong model capability signal. - xAI
Grok Build 0.1moves into the agentic coding watchlist: OpenRouter and third-party indexes describe it as a 256K-context coding model for software engineering agents;@xiaohualso noted that X Premium now supports Openclaw, Opencode, and Hermes OAuth login using Grok usage. Qwen3.7-Maxremains a practical testing priority: both OpenRouter and Qwen Cloud provide API access, and LMArena still listsqwen3.7-max-previewat rank 14.- Video leaderboards did not materially change:
Dreamina Seedance 2.0 720pandHappyHorse-1.0still lead Artificial Analysis T2V/I2V, whileGemini Omnistill does not appear on that leaderboard.
Must Watch Today
1. Claude Mythos Preview / Project Glasswing
- Event type: official research update + model capability signal.
- Official update date: Anthropic published the Project Glasswing initial update on 2026-05-22.
- Capability change: Anthropic says Mythos Preview scanned 1,000+ open-source projects and identified an estimated 23,019 vulnerability items, including 6,202 high- or critical-severity items. Anthropic's X account continued to highlight Project Glasswing and Claude Mythos Preview's vulnerability discovery capability on 2026-05-23.
- Availability: Mythos Preview is still not a public model. It is available to partners through Project Glasswing for defensive security research.
- Impact: this is not a normal product launch, but it directly matters for whether models can perform long-horizon, high-value security research tasks. Continue tracking whether Mythos gets broader controlled access or whether its capabilities move into the Opus family.
- Sources: Anthropic Project Glasswing update, Project Glasswing, Anthropic X
2. Grok Build 0.1 / Grok Agentic Coding Access
- Event type: model availability + X access-ecosystem signal.
- Observed timing: OpenRouter and model indexes show a mid-May 2026 launch;
@xiaohuposted that X Premium members can now authorize Openclaw, Opencode, and Hermes directly, using X Premium Grok usage to power those products. - Capability notes: third-party indexes describe
Grok Build 0.1as a coding model for agentic software engineering workflows, supporting 256K context, tool use, structured outputs, and related coding-agent features. - Availability: xAI Console / OpenRouter / third-party Grok coding-agent ecosystem. The X Premium OAuth path needs account-level verification.
- Impact: if the user already has X Premium or SuperGrok, the cost structure of powering coding agents through Grok usage may be worth evaluating. Public benchmark coverage is still thin, so do not assume it beats Codex, Claude Code, or Qwen3.7-Max.
- Sources: Xiaohu X, OpenRouter Grok Imagine / xAI model pages, LLM Reference Grok Build, KuCoin xAI Grok Build note
3. Qwen3.7-Max
- Event type: availability continuation + leaderboard confirmation.
- Official release date: Alibaba Cloud Summit announcement on 2026-05-20; Qwen's official X account pinned
Qwen3.7-Maxon 2026-05-21. - Last 24-hour change: no new official upgrade found, but OpenRouter and Qwen Cloud API access remain live, and Qwen X still preserves the OpenRouter availability signal.
- Availability: OpenRouter lists
qwen/qwen3.7-max, releasedMay 21, 2026; Qwen Cloud provides OpenAI-compatible API examples. - Capability notes: positioned for agent-centric workloads, especially coding, office/productivity, and long-horizon autonomous execution; Qwen Cloud lists function calling, cache, structured outputs, batches, and web search support.
- Leaderboard status: LMArena Text Overall lists
qwen3.7-max-previewat rank 14, Elo1475 ±10, 3,741 votes, Preliminary. - Impact: still one of the most important closed APIs to test. Focus on tool-call stability, long-context cost, and multi-file coding-agent behavior.
- Sources: Qwen X, OpenRouter, Qwen Cloud, Alibaba Cloud, LMArena
4. NetEase Youdao Confucius4 / Confucius4-TTS
- Event type: ongoing open model card/repository confirmation.
- First signal:
@xiaohuposted the Confucius4 two-model signal around 2026-05-22. - Current status: the
Confucius4Hugging Face page is still visible, based onQwen3.5-27Band Apache 2.0; theConfucius4-TTSGitHub repository still says code and model weights are in preparation. - Quantitative drift: in this run,
Confucius4-TTShad 77 GitHub stars and 2 issues; the Hugging Face page had 15 likes. Attention is increasing, but engineering availability is still incomplete. - Impact: sample the multimodal model on visual math reasoning. Keep TTS in watch mode until code and weights are complete.
- Sources: Confucius4 Hugging Face, Confucius4-TTS GitHub, Xiaohu X
5. Cohere Command A+
- Event type: distribution continuation after official release.
- Official release date: 2026-05-20.
- Last 24-hour change: Cohere's X page still shows Microsoft Azure's post saying Command A+ is available in Microsoft Foundry as a Managed Compute offer; no new model-spec change found.
- Capability notes: Cohere calls it its fastest and most powerful Command model, with an MoE architecture,
218B total / 25B active, 128K input context, and support for text, images, tool use, reasoning, and 48 languages. - Impact: keep it in the enterprise private deployment, RAG, tool-use, and long-document workflow candidate pool. Priority is lower than today's Mythos capability signal and Qwen3.7-Max hands-on testing.
- Sources: Cohere Blog, Cohere X
Arena Watch
- LMArena Text Overall: leaderboard date is still 2026-05-18.
gemini-3.5-flashis rank 9, Elo1480 ±8;qwen3.7-max-previewis rank 14, Elo1475 ±10; both are Preliminary. - Artificial Analysis T2V with audio:
Dreamina Seedance 2.0 720pranks 1st with Elo1214;HappyHorse-1.0ranks 2nd with Elo1213;Veo 3.1ranks 3rd with Elo1101. - Artificial Analysis I2V with audio:
Dreamina Seedance 2.0 720pranks 1st with Elo1185;HappyHorse-1.0ranks 2nd with Elo1169;Veo 3.1 Fastranks 3rd with Elo1094. - Watch item:
HappyHorse-1.0is still listed as Alibaba-ATH, ReleasedApr 2026, APIComing soon, with no formal model card or API documentation found;Gemini Omnistill does not appear on the leaderboard. - Sources: LMArena Text, Artificial Analysis T2V, Artificial Analysis I2V
X Watch
@xiaohu: the latest visible high-engagement item is X Premium authorization for Openclaw, Opencode, and Hermes, using Grok usage to power those products. This is an AI coding-agent access-ecosystem signal, not a new model announcement.@Alibaba_Qwen: still pinsQwen3.7-Maxand keeps the OpenRouter launch repost visible.@cohere:Command A+remains pinned; Microsoft Azure Foundry distribution signal remains visible.@GoogleDeepMind: still pinsGemini Omni; the 2026-05-23 Singapore AI deployment partnership is not a new model release.@OpenAI: the pinned item remains the planar unit distance problem model capability breakthrough; the 2026-05-22 Codex Thursday update is Codex/product capability, not a model launch.@AnthropicAI: the 2026-05-23 Project Glasswing update is today's highest-value model capability signal because it points to Claude Mythos Preview's security research capability.- X search: this run loaded, but the results were noisy and contained many non-model items. The report prioritizes account pages and official sources.
Pending Confirmation
Claude Mythos Preview: no public access plan yet. Continue tracking broader controlled access, pricing, API availability, or capability migration into the Opus line.Grok Build 0.1: confirm official xAI docs, CLI access path, X Premium authorization boundary, and real benchmarks. Third-party indexes justify monitoring, but not a conclusion.Gemini OmniAPI: Google still says API access is coming in the next few weeks; continue checking Gemini API, AI Studio, Vertex AI, and Flow.Gemini Omnileaderboard: no independent Artificial Analysis video leaderboard entry found yet.HappyHorse-1.0: leaderboard lists Alibaba-ATH, ReleasedApr 2026, APIComing soon, but there is still no formal model card or API documentation.Confucius4-TTS: GitHub repository is public, but code and weights are still in preparation.- OpenAI/Google/Meta/Mistral/Qwen: no new major flagship model announcement found in the last 24 hours.
Next Steps
- Put
Qwen3.7-Max,Grok Build 0.1, andCommand A+into the same coding agent / multi-file edit benchmark group. - Track
Claude Mythos Previewas a capability signal only, not an integration candidate; it remains a controlled security research model. - Run a small visual math sample set on
Confucius4; wait forConfucius4-TTScode and weights before testing. - Continue watching
Gemini OmniAPI access andHappyHorse-1.0API availability; either one would change video-model testing priority.