OpenAI vs Anthropic Memo · Zhipu GLM-5.1 Open-Source · Google Gemma 4

12-hour AI industry digest

Top Stories

1. OpenAI Slams Anthropic to Investors: “Substantially Ahead” in Compute

Date: April 9
Source: CNBC, Bloomberg

OpenAI sent a memo to investors slamming main competitor Anthropic, characterizing it as “operating on a meaningfully smaller curve” and compute-constrained. OpenAI stated it plans to have 30 gigawatts of compute by 2030, while expecting Anthropic to have roughly 7-8 gigawatts by end of 2027.

The memo noted: “Even at the high end of that range, our ramp is materially ahead and widening.” Both companies are collectively valued at over $1 trillion and gearing up for possible IPOs this year, while trying to convince investors they have sustainable business models.

Impact: OpenAI-Anthropic competition extends from technical level to investor relations warfare, reflecting market positioning battles ahead of IPOs.

2. Zhipu AI Open-Sources GLM-5.1: 744B Parameter MoE Model Free to Use

Date: April 7-8
Source: whatllm.org, Planet News

Zhipu AI released its flagship model GLM-5.1 as open-source while raising API prices by 10%. This is a 744-billion-parameter mixture-of-experts (MoE) model, activating 40 billion parameters per forward pass, with 200K context window. On SWE-Bench Pro (expert-level real-world software engineering), it reportedly defeated both Claude Opus 4.6 and GPT-5.4.

The model is released under MIT license, with usage cost being only electricity expenses. This is the company’s second price increase this year, following a comprehensive reform in February that raised coding subscription plan rates by over 30%.

Impact: Chinese AI enterprises are exploring “open-source + commercialization” dual-track strategy to maintain advantage in global competition.

3. Google Releases Four Gemma 4 Models in One Day: Fully Open-Source

Date: April 1
Source: whatllm.org

Google released four Gemma 4 series models on April 1, all under Apache 2.0 open-source license:

• Gemma 4 27B - Text + Image + Audio multimodal
• Gemma 4 26B-A4B - Mixture-of-experts architecture
• Gemma 4 E2B / E4B - Efficient inference versions

All models are free for self-hosting, marking Google’s significant investment in open-source AI.

Impact: Google counters Meta and Zhipu AI’s open-source model offensives through comprehensive open-source strategy.

Model & Technical Progress

4. Alibaba Releases Qwen 3.6-Plus: Agentic Capabilities Upgrade

Date: April 2
Source: whatllm.org

Alibaba Cloud released Qwen 3.6-Plus, a text model supporting agentic capabilities. The model uses open-source licensing, with API pricing at approximately $0.28 per million tokens, competitive on price-performance.

Highlight: Qwen series continues iteration, strengthening agentic capabilities to address enterprise automation demands.

5. Microsoft Releases MAI-Transcribe-1: Azure Speech-to-Text Service

Date: April 2
Source: whatllm.org

Microsoft released MAI-Transcribe-1, a proprietary speech-to-text model available through Azure. This model represents Microsoft’s latest advances in enterprise speech recognition.

Impact: Microsoft continues expanding its Azure AI services portfolio, deeply binding with enterprise customers.

Industry & Business

6. AI Model Price Divergence: From Free to $125/Million Tokens

Date: April 7-8
Source: whatllm.org

AI model releases in early April 2026 show unprecedented price divergence:

• Free: GLM-5.1 (MIT license), Gemma 4 series (Apache 2.0)
• Low-cost: Qwen 3.6-Plus (~$0.28/M)
• High-cost: Claude Mythos ($25/M input, $125/M output)

Price differences no longer just about capability, but about control. The industry’s most powerful models are being built faster than anyone can agree on who should use them.

Impact: AI industry shows philosophical split—open vs. closed, free vs. expensive coexist.

7. Tencent Cloud AI Service Price Increase 5%: Effective May 9

Date: April 9
Source: xix.ai

Tencent Cloud announced that due to strong global demand and rising hardware costs, core AI services, containers, and big data products will increase prices by 5% effective May 9, 2026. Existing customer contracts are unaffected until renewal.

Impact: Global AI infrastructure cost increases beginpass through to end-user service prices.

Trend Watch

Key Trends in Past 12 Hours

1. Open vs. Closed Philosophical Split: 8+ models released in 7 days in early April, with 5 open-weight models coexisting with closed commercial models, reflecting fundamental industrydivergence on “who should use AI.”

2. IPO Investor Relations Warfare: OpenAI and Anthropic publicly clash through investor memos ahead of IPOs, reflecting intensifying competition in valuation and business model positioning.

3. Chinese AI Dual-Track Strategy: Zhipu AI’s “open-source model + price increase API” strategy represents new thinking from Chinese AI enterprises—building ecosystem through open-source, achieving sustainability through commercialization.

4. Multimodal Becomes Standard: All Google Gemma 4 series support text+image+audio, with multimodal capabilities moving from high-end closed models to open-source models.

5. Compute Arms Race: OpenAI 30 GW vs Anthropic 7-8 GW compute planning shows AI infrastructure competition entering new magnitude.

Upcoming Watch

• OpenAI GPT-5.5 (codename Spud): Expected release in coming weeks
• DeepSeek V4: In gray testing, new vision version and “expert mode”launching soon
• Anthropic Claude Mythos: 50-company limited access, no public release date
• EU AI Act Implementation: Shifting from draft to execution phase, may affect global AI deployment

Briefing generated: 2026-04-10 06:00 (Asia/Shanghai)
Coverage period: 2026-04-09 18:00 — 2026-04-10 06:00
Data sources: Public news reports, industry analysis, official announcements