- 29
- May
May 2026 was one of the busiest months in AI all year — within a single month, three major labs changed their default model at once: OpenAI shipped GPT-5.5 Instant (May 5), Google unveiled Gemini 3.5 Flash + Omni + Spark at I/O 2026 (May 19), and Anthropic released Claude Opus 4.8 with Dynamic Workflows (May 28-29). The month closed with reports that Anthropic is negotiating a $900 billion funding round that could overtake OpenAI as the most valuable AI startup in the world.
This article is an "index of May's AI news" — everything in one post, with links to read the full detail elsewhere. Too busy all month? Read this one and you won't miss anything. Most importantly, it closes with how Thai organizations and ERP teams should adapt.
May 2026 by the Numbers
Contents
Timeline — the month's key events
The month at a glance, by date. Missed a week? Jump to that row.
| Date | Event | Details |
|---|---|---|
| May 5 | OpenAI GPT-5.5 Instant | Shipped GPT-5.5 Instant as ChatGPT's new default across every tier (Free → Enterprise), replacing GPT-5.3 Instant — tighter, more accurate answers |
| May 12 | Anthropic $30B round talks | Bloomberg reported Anthropic is in talks to raise >$30B at a >$900B valuation — poised to pass OpenAI ($852B in March) |
| May 19 | Google I/O 2026 | Unveiled Gemini 3.5 Flash (fast/cheap for agents+coding), Gemini Omni (world model that outputs video), and Gemini Spark (personal agent) + Android XR + AI Search |
| May 22 | Anthropic round nearing close | Bloomberg reported the round could close "as soon as next week" — Sequoia, Dragoneer, Altimeter, Greenoaks among co-leads |
| May 28 | Claude Opus 4.8 | Anthropic released Opus 4.8 as the default for premium users — SWE-bench Pro 69.2%, 4x fewer missed code flaws, Dynamic Workflows, Effort Dial |
| May 29 | Claude Code Dynamic Workflows | Claude Code enabled Dynamic Workflows by default (tens to hundreds of subagents) + plugins in web chat + plugin marketplaces |
OpenAI — GPT-5.5 Instant as new default
On May 5, OpenAI shipped GPT-5.5 Instant as the default ChatGPT model across every tier — Free, Go, Plus, Pro, Business, Enterprise, and Edu — replacing GPT-5.3 Instant. The pitch: tighter, more accurate answers with less "yapping," plus better reasoning and multimodal scores.
| GPT-5.5 Instant highlights | Details |
|---|---|
| Default on every tier | All ChatGPT users get it automatically — no manual selection |
| Tighter answers | Less filler, more direct and usable responses |
| Benchmarks | AIME 2025 (math) 81.2 and MMMU-Pro (multimodal) 76 |
| Agent direction | Codex reportedly can control a Mac even when the screen is locked — a notable autonomous-agent step |
Recommendation: if your team uses ChatGPT via the API or prompts tuned to an older model, re-test against GPT-5.5 — the shorter answer style may affect workflows that expect long outputs. See our earlier review of GPT-5.4.
Google I/O 2026 — Gemini 3.5 Flash / Omni / Spark
Google I/O 2026 (May 19) was the mid-month highlight, with several launches centered on "AI agents + multimodal + search."
| What's new | What it is | Why it stands out |
|---|---|---|
| Gemini 3.5 Flash | Fast, cost-efficient model for AI agents and coding | Near-flagship intelligence at Flash speed |
| Gemini Omni | New model family taking image/audio/video/text and generating video grounded in real knowledge, easily edited | Combines reasoning + creation (a world model) |
| Gemini Spark | A "personal agent" in the Gemini app that acts on your behalf — reasoning across connected apps | Starts with Google AI Ultra (US) |
| Gemini 3.5 Pro | Already used internally; rolling out next month | The flagship reasoning tier |
The most telling launch is Gemini Spark — a personal agent that works across apps, matching the same direction as Agentic AI every lab is chasing (Claude Dynamic Workflows, OpenAI Codex). May made it clear that "the agent" is now the main battleground.
Anthropic — Claude Opus 4.8 + $900B round
Anthropic closed out the month with two big stories back to back:
1. Claude Opus 4.8 (May 28)
Claude Opus 4.8 launched as the default model for premium users everywhere — stronger at coding and long-running agentic work, same price at $5/$25 per 1M tokens. Key new features:
| Feature | Details |
|---|---|
| SWE-bench Pro 69.2% | Agentic coding, up from 64.3% on Opus 4.7 |
| 4x fewer missed flaws | Less likely to let code flaws pass unnoticed — good for code review |
| Dynamic Workflows | In Claude Code — fans out across tens to hundreds of subagents + self-checks |
| Effort Dial + Fast Mode | Adjustable thinking effort + Fast Mode ~2.5x faster, 3x cheaper |
Read the full breakdown in our Claude Opus 4.8 review for developers.
2. A reported $900B funding round (Bloomberg)
Bloomberg reported mid-to-late month that Anthropic is in talks to raise more than $30 billion at a pre-money valuation above $900 billion, with Sequoia, Dragoneer, Altimeter, and Greenoaks each expected to co-lead with roughly $2 billion.
Note: the deal is not finalized (no term sheet signed as of late May, per reports) — but on the reported terms, Anthropic would surpass OpenAI ($852B in March 2026) as the world's most valuable AI startup, roughly a 15x valuation increase in about 14 months.
Money & structure — the race levels up
May showed that the AI race is no longer just "whose model is smartest." It now spans capital, compute infrastructure, and the shift to agents.
- Record capital inflows — the reported $900B Anthropic round, alongside reports of revenue growing many-fold year over year, signals real enterprise-AI demand (see context in Anthropic and the enterprise AI market)
- The battle moved to "the agent" — Gemini Spark, Claude Dynamic Workflows, and OpenAI Codex (controlling a locked machine) all push AI to "do the work," not just answer questions
- Compute is the big bet — reports of large compute deals among major players reinforce that "who has more GPUs/compute" is a decisive factor in the next round
Key insight: for Thai organizations, the takeaway isn't "chase every headline." It's that models and prices change fast, so tying your core systems to a single vendor is risky. Design so you can switch or add providers, and keep critical data in your own systems.
Impact on Thai organizations — what to adjust
May's news affects Thai organizations across five dimensions, especially those using ERP or starting to connect AI to their operations.
| Dimension | Impact | What to do | Urgency |
|---|---|---|---|
| Default models changed | GPT-5.5, Gemini 3.5, Opus 4.8 change answer behavior | Re-test existing prompts/integrations on the new models | This quarter |
| Agents do the work | Spark/Workflows/Codex can genuinely act across apps | Find repetitive use cases for AI but control data access | This quarter |
| AI cost | Dynamic Workflows is token-heavy; tier pricing changes often | Set budget caps + pick effort/fast mode per task | Now |
| Vendor risk | Capital and value concentrate in a few players | Multi-vendor + DR plan + don't lock workflows to one provider | This quarter |
| Security / governance | Agents access more data = bigger risk surface | Review security policy + AI governance before letting agents act | Now |
The ERP view: AI changed defaults all month — models come and go. But your ERP system is the foundation that doesn't change. Whatever AI you connect, your accounting, inventory, procurement, and finance data must stay accurate and auditable. Smart organizations get the ERP foundation solid first, then layer AI on top — Saeree ERP is developing an AI Assistant in this direction (still in development, not yet generally available).
6 lessons from May 2026
| # | Lesson | Evidence |
|---|---|---|
| 1 | 3 labs changed defaults at once | GPT-5.5, Gemini 3.5 Flash, Claude Opus 4.8 in one month |
| 2 | The battleground moved to "agents" | Gemini Spark + Claude Dynamic Workflows + OpenAI Codex |
| 3 | Capital concentrates in a few players | Anthropic's reported $900B round poised to pass OpenAI |
| 4 | Price/speed move fast | Fast Mode 3x cheaper; Gemini 3.5 Flash emphasizes cheap/fast |
| 5 | "Honest" models matter more | Opus 4.8 lets 4x fewer code flaws slip — good for production |
| 6 | The data foundation is what lasts | Models change monthly, but ERP/enterprise data must stay stable |
May 2026 taught us that AI changes its defaults every month, and the competition has spread from "intelligence" to "capital + agents + compute." The Thai organizations that win aren't the ones chasing every headline — they're the ones that choose vendors wisely, use AI to augment people, and build a solid ERP foundation first, then layer AI on top.
- Sureeraya Limpaibul, Managing Director, Grand Linux Solution
Article index — read the full detail
This is the overview. Want to go deep on a topic? Follow the dedicated articles below (some topics will be added next month).
AI Models / Products
- Claude Opus 4.8 (May 28) — Full review: SWE-bench Pro 69.2%, Dynamic Workflows, Effort Dial, Fast Mode
- Claude Opus 4.7 (prior version) — Coding +13%, Vision 3x, xhigh effort, /ultrareview
- GPT-5.4 (OpenAI, prior) — 1M Token Context, Computer Use, Agentic Workflows
- GPT-5.5 Instant — (coming soon)
- Google I/O 2026 — Gemini 3.5 / Omni / Spark — (coming soon)
- ChatGPT vs Claude vs Gemini — comparing the three
Industry direction
- What is Agentic AI — AI that does the work
- Claude Managed Agents — read more
- Codex vs Claude Code — comparing coding agents
- Anthropic and the enterprise AI market — read more
- China's tech strategy — read more
- AI replacing vs augmenting people — read more
- AI layoffs Q1 2026 — read more
ERP + organizational basics
- What is ERP — read more
- ERP implementation — read more
- AI Governance — read more
- Security basics — read more
- Disaster Recovery — read more
Previous AI roundups
- April — 10 days of Songkran — 10 Days of AI You Missed (9-18 April 2026)
- March 2026 — March AI news roundup
References
- OpenAI — Introducing GPT-5.5
- TechCrunch — OpenAI releases GPT-5.5 Instant as default ChatGPT model (May 5, 2026)
- Google — 100 things announced at I/O 2026
- 9to5Google — Everything Google announced at I/O 2026 (May 19, 2026)
- CNBC — Google debuts new AI models, personal AI agents
- Anthropic — Introducing Claude Opus 4.8
- Bloomberg — Anthropic In Talks to Raise $30 Billion at $900 Billion Valuation
- Bloomberg — Anthropic to Close Over $30 Billion Round as Soon as Next Week
