GLM-5.1 Launches — 94.6% of Claude Opus 4.6 Coding at $3/month
Z.ai — the lab formerly known as Zhipu AI — has launched GLM-5.1, and the coding benchmark numbers deserve a second look. The model scored 45.3 on Claude Code's own coding evaluation framework, compared to Claude Opus 4.6's 47.9. That's 94.6% of Opus-level coding performance, achieved in roughly six weeks of iteration from GLM-5's baseline score of 35.4. A 28% jump in six weeks is a fast-moving target.
What makes this launch particularly notable is the pricing. GLM-5.1 is available to GLM Coding Plan subscribers starting at $3/month on promotion, $10/month at standard rates. For developers running heavy coding workloads who've been eyeing Claude Opus 4.6 but balking at the cost, this is a credible alternative worth benchmarking against your actual use cases.
The technical setup is straightforward: existing API users switch to the new model simply by changing the model name in their API call to `glm-5.1`. No migration headaches, no new tooling. The fact that Z.ai tested inside Claude Code's own evaluation environment — and published the comparison transparently — adds a layer of credibility to the numbers that self-reported benchmarks often lack.