GLM-5.2 turned a routine-looking model release into a market event. On June 15, Zhipu AI's Hong Kong-listed shares jumped as much as 48% intraday to HK$1,620 and closed up 32.8% at HK$1,457.
The stock is now up roughly 820% since Zhipu's IPO in early January. JPMorgan raised its price target to HK$1,400 and kept an overweight rating, and Bank of America started coverage at buy. A version bump does not usually move a company like that. This one did.
The dollar figures put it in perspective. The day before GLM-5.2 shipped, Z.AI's listed shares were worth around $60 billion. The launch added close to $20 billion in a single session and lifted the company to roughly $80 billion, about the size of Airbnb. The easiest yardstick is Coinbase, worth about $42 billion in mid-June 2026: GLM-5.2 took Z.AI from roughly one and a half Coinbases to nearly two in a single day. A coding-model release is now moving a company the size of a US tech blue chip.
Why the stock moved
GLM-5.2 is Z.AI's new flagship coding model, and two things made its launch a market story rather than a footnote.
First, it is open source, released under the MIT license with a 1-million-token context window and a focus on long-horizon agentic coding. It extends a lineage that has been closing on the closed frontier: GLM-5 scored 77.8% on SWE-bench Verified, and each release since has narrowed the gap. Because the weights are downloadable, the capability cannot be revoked.
Second, the timing. GLM-5.2 arrived the same weekend a US export-control order forced Anthropic to disable its two most powerful models, Claude Fable 5 and Mythos 5, for every user. One frontier vendor went dark by government directive. An open model with a frontier-class context window showed up at roughly a tenth of the price of Anthropic's top Claude Code and Max tiers. Investors read that as Chinese open models stepping into the gap, and re-rated the company that ships them.
Whether the rally holds is a question for the market. The more useful question for a developer is narrower: is the model behind it actually good on your code? You do not have to buy the stock to find out.
Try GLM-5.2 free on Token Station
GLM-5.2 is live on Token Station as glm/glm-5.2, with the full 1M-token context, at Z.AI's list price and zero markup: $1.40 per million input tokens and $4.40 per million output. One thing to budget for: thinking is always on, so reasoning tokens bill as output.
It is free to start. Register, with no card and no Z.AI account or Coding Plan subscription to set up. Your first top-up adds up to $50 in bonus credit. Point the coding tools you already use at glm/glm-5.2 and run your real work through it.
Claude Code
Claude Code reads its model and endpoint from environment variables. Route every tier through Token Station to GLM-5.2:
export ANTHROPIC_BASE_URL="https://models.bytefuture.ai"
export ANTHROPIC_AUTH_TOKEN="gw-YOUR_TOKEN_STATION_KEY"
export ANTHROPIC_DEFAULT_OPUS_MODEL="glm/glm-5.2"
export ANTHROPIC_DEFAULT_SONNET_MODEL="glm/glm-5.2"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="glm/glm-5.2"
export CLAUDE_CODE_SUBAGENT_MODEL="glm/glm-5.2"
claudeCodex
Configure Token Station as the provider and make GLM-5.2 the model:
mkdir -p ~/.codex
cat > ~/.codex/config.toml <<'EOF'
model = "glm/glm-5.2"
model_provider = "token_station"
[model_providers.token_station]
name = "token_station"
base_url = "https://models.bytefuture.ai/v1"
env_key = "TOKEN_STATION_API_KEY"
wire_api = "responses"
EOF
export TOKEN_STATION_API_KEY="gw-YOUR_TOKEN_STATION_KEY"
codexOpenClaw
Register Token Station as a provider and set GLM-5.2 as the default model:
{
"models": {
"mode": "merge",
"providers": {
"token-station": {
"baseUrl": "https://models.bytefuture.ai/v1",
"apiKey": "${TOKEN_STATION_API_KEY}",
"api": "anthropic-messages",
"models": [
{
"id": "glm/glm-5.2",
"name": "GLM-5.2 (Token Station)",
"contextWindow": 1000000,
"maxTokens": 131072
}
]
}
}
},
"agents": {
"defaults": {
"model": { "primary": "token-station/glm/glm-5.2" }
}
}
}One key, the harness you already run, and the model the market just re-rated. If GLM-5.2 holds up on your repository, a free signup is all it costs to find out. If it does not, you change one line of config and move on.
Start here: models.bytefuture.ai