✳ the wire · intel
GLM-5.5 may skip 5.3 and 5.4 entirely — August target, open weights, 1M context
An email when GLM-5.5 answers on a public API, about 15 minutes after members, plus every frontier launch · no card
Fresh leaks tighten the picture on Zhipu's next flagship. The version number appears to be settling on 5.5 with 5.3 and 5.4 skipped outright, the target remains August, the 1M-token context window carries over from GLM-5.2, weights stay open, and the design focus is long-horizon autonomous coding agents rather than chat quality.
Parameter count is where the reporting diverges and you should be careful. The JPMorgan note relayed by Reuters in June said 'more than 1T'; community leaks circulating this week say 3T+. GLM-5.2 is 753B. We are carrying the range, not picking a number — a 3× spread in the only hard spec anyone claims is a sign the leak chain has more repetition than sourcing in it.
Every wire item is source-verified before it posts and graded: confirmed · strong · reported · rumor. Subscribers see the primary source on each one, the reasoning behind its grade, and the full analysis — plus the launch alerts themselves, which fire the minute a model answers rather than whenever the feed catches up.
