Z.ai's GLM-5.3 scores 60 on AI index, matching Kimi K3

Z.ai's GLM-5.3 scored 60 on the Artificial Analysis Intelligence Index, matching Moonshot AI's Kimi K3 and trailing Anthropic's Claude Opus 5 by three points, while undercutting both on price. The independent evaluator's reading, published Aug. 18, places the Chinese lab's newest reasoning model eighth among 181 models in its comparison class, well above the 35 median.

"GLM-5.3 shows significant improvement in coding and agent capabilities," CMBI said in a note reiterating its Buy rating on Z.AI (02513.HK). The model's gains come entirely from post-training rather than a new pretraining run, Z.ai said.

Z.ai released GLM-5.3 on Aug. 14 using the same base model as GLM-5.2, with a month of scaled reinforcement learning on long-horizon task environments. That approach moved Terminal-Bench 3.0 from 4.6 to 28.3 and DeepSWE v1.1 from 46.2 to 66.9. On Z.ai's API, GLM-5.3 costs $1.40 per million input tokens and $4.40 per million output tokens, against $3.00 and $15.00 for Kimi K3 on Moonshot's platform. Per Intelligence Index task, that works out to $0.68 for GLM-5.3 versus $0.84 for Kimi K3 and $2.34 for Claude Opus 5.

The API opened for access Aug. 18 with pricing unchanged from GLM-5.2, and the model has been integrated into Z.ai's ZCode coding platform and GLM Coding Plan. Model weights are scheduled to be open-sourced Aug. 28, two weeks after launch, once safety evaluation and hardening are complete. Z.AI shares fell 4.3% on the day, with short selling of $639.41 million at a 5.698% ratio.

The parity with Kimi K3 is the headline comparison. Kimi K3, released July 16, remains the top-scoring open-weights model in Artificial Analysis's rankings, a position GLM-5.3 now shares in score if not in license. Moonshot opened Kimi K3's weights under a revenue-tiered license in July; GLM-5.3 is listed as proprietary for now, with Artificial Analysis recording 753 billion parameters for the model.

Claude Opus 5, released July 24, still leads the index at 63. The three-point gap between GLM-5.3 and the leader is the distance a fast post-training cycle did not close, against a frontier that has itself moved since the spring.

The cost advantage is where the two 60-scorers diverge sharply. GLM-5.3 reaches its score at the lowest cost per task of the three, though it is also the most verbose of the group, generating 170 million output tokens across the evaluation suite against a 72 million median in its class. That verbosity could offset some of the per-token savings in real-world coding workloads.

The model requires thinking to be enabled, with three effort levels, and Z.ai warns that applications still calling it with thinking disabled will fail until migrated. The weights release next Friday would put GLM-5.3 alongside Kimi K3 as an open-weights option at the 60 mark on the index, at roughly half the per-token price.

Z.AI shares, which fell 4.3% on the day, trade against intensifying competition in open-weights models. The open-sourcing of GLM-5.3 next week could accelerate developer adoption and strengthen the company's position against Moonshot, Anthropic, and OpenAI's GPT-5.6 Sol, which the model now matches on the index. CMBI's reiterated Buy rating suggests the selloff may be overdone, but the three-point gap to Claude Opus 5 shows the frontier has not stopped moving.

This article is for informational purposes only and does not constitute investment advice.