Zhipu AI Raises $4B at $128B Valuation — GLM-5.2 Goes Free

Why this matters now

Chinese AI company Zhipu AI — the developer behind GLM-5.2 — is raising approximately $4 billion through a share sale, after its stock surged nearly 1,500% since its Hong Kong listing in January 2026. The company’s market capitalization has topped HK$1 trillion (~$128 billion), making it one of the most valuable AI companies globally by market cap — behind only OpenAI, Anthropic, and a handful of US hyperscalers.

This isn’t just a financial story. Zhipu has simultaneously made GLM-5.2 free via API — an open-weight, MIT-licensed model that scores within 1% of Claude Opus 4.8 on a key agentic benchmark at roughly 20% of the cost. The combination of a $128B valuation and a free frontier-class model creates a dynamic the US AI industry hasn’t faced before: a well-capitalized Chinese competitor that isn’t trying to maximize API revenue.

The financials

JPMorgan raised its 2026–2030 revenue forecast for Zhipu by 7–16% following GLM-5.2’s launch, citing the model as evidence of Zhipu’s “growing pricing power” — notably, the company removed the discounts it had offered on earlier model API tiers. The bank expects Zhipu to turn profitable in 2028, a revision from its previous net loss forecast.

GLM-5.2: the model that changed the narrative

GLM-5.2, released June 13 under an MIT license, is the engine behind Zhipu’s valuation surge. Key specs from our earlier coverage:

• 1M token context window — 8x Claude Sonnet 5, 4x GPT-5.6 Sol
• Dual thinking-effort system — fast mode for simple queries, deep mode for hard reasoning
• 131K max output tokens — useful for long-form generation
• MIT license — full commercial freedom
• Pricing: ~$0.73–1.40/M input, ~$2.29–4.40/M output (before the free tier)

The benchmark positioning is what caught the market’s attention. GLM-5.2 scores within 1 percentage point of Claude Opus 4.8 on agentic benchmarks while costing roughly one-fifth as much. For enterprise teams watching their API bills climb with Sonnet 5 and Opus 4.8 usage, that value proposition is hard to ignore.

The free API: a strategic weapon

Zhipu’s decision to offer GLM-5.2 free via cloud API is the most aggressive pricing move in the current market. The calculus:

1. Developer adoption first, revenue second — Give the model away free, embed it in workflows, then monetize through scale, fine-tuning, and enterprise services
2. Undercut every competitor — At $0/M tokens, Zhipu is cheaper than Hy3 ($0.20), Luna ($1/$6), Mistral Medium 3.5 ($1.50/$7.50), and every other option
3. Bypass export controls — Open-weight + free API means developers anywhere can access it, regardless of US government restrictions on frontier models

This is the same playbook DeepSeek used, but Zhipu has better timing: US frontier models are increasingly government-gated (Fable 5, Mythos 5, GPT-5.6 Sol preview), while Chinese open-weight models flow freely.

Where Zhipu sits vs US competitors

Zhipu at $128B is now worth ~9x Mistral and closing the gap with the US frontier. The free API strategy suppresses short-term revenue in exchange for market share — a bet that developer mindshare today converts to enterprise contracts tomorrow.

What this means for builders

If you’re running a multi-provider AI stack, Zhipu’s GLM-5.2 free tier changes the routing math:

• For extraction, classification, and chat — GLM-5.2 free is cheaper than any paid alternative, including Luna and Hy3
• For agentic coding — GLM-5.2 is competitive with Sonnet 5 and Terra on quality while being free on API calls. The trade-off is latency and throughput (free tier rate limits)
• For long-context work — GLM-5.2’s 1M context at free pricing is unmatched
• For self-hosted deployment — MIT license means no restrictions, and you can run it on your own hardware

The catch: it’s a Chinese company operating under Chinese law. For teams with data sovereignty requirements, the self-hosted option (MIT weights) is the safe path. For teams that can use the API, the pricing is hard to beat.

Decision framework

Use GLM-5.2 free tier if:

• Your workload is high-volume extraction, classification, or chat where cost is the primary constraint
• You need 1M+ context and can’t justify Gemini 3.1 Pro pricing
• You want to test an open-weight model before committing to self-hosted deployment

Use GLM-5.2 self-hosted if:

• Data sovereignty prevents you from using Chinese API services
• You want a zero-cost inference option on your own hardware
• You need the MIT license for commercial flexibility

Skip GLM-5.2 if:

• Your primary use case is autonomous coding that requires top-tier accuracy — Opus 4.8 or Sonnet 5 still lead there
• You have regulatory restrictions on using Chinese AI models in your organization
• You need guaranteed throughput and SLA commitments — the free tier has no guarantees

The bottom line: Zhipu’s $4B raise and $128B valuation confirm that open-weight AI from China is not a niche story — it’s a structural shift in the AI market. A $128B company giving away frontier-class models for free changes the competitive dynamics for every other model provider, and it makes GLM-5.2 the default cost-optimizer in any multi-provider routing strategy.

Related reading

• GLM-5.2: 1M Context and IndexShare Sparse Attention for Coding
• The Agent Model Price Guide, July 2026
• Tencent Hy3: A 295B Open MoE Model Built for Agents
• How to Set Up Multi-Provider AI Routing in Laravel

Sources

• Bloomberg/Yahoo — AI Firm Zhipu to Sell $4 Billion of Shares After 1,500% Rally
• SCMP — Zhipu AI market cap tops HK$1 trillion
• WSJ — China’s Zhipu AI Shares Surge on $4 Billion Fundraising
• JPMorgan — Zhipu AI revenue forecast upgrade
• SaasRise — Zhipu AI’s Free GLM-5.2 Stirs US Investor Anxiety

About the author: Charles Jasthyn De La Cueva is an Admin Officer at PSU’s Quality Assurance Office and the builder behind ParSU-EDMS / QAOSYS. He writes about AI tools, infrastructure, and practical agent deployment.