Z.ai (Zhipu) · Benchmarks

GLM 5.3 benchmarks

GLM 5.3 has 6 published scores in this dataset, from a Z.ai (Zhipu) release dated 2026-08-14. Its composite index is 82.7, which places it 14th of 22 ranked models and 9th among 15 open-weight models. Its strongest area is agentic (index 89.3) and its weakest is coding (index 78.3). Its best placing is on LMArena Text Elo, where its 1481 ranks 6th of 21 models with a published score.

Open weightsOfficial page

Category profile

Index 0–100: each score relative to the best published score on that test.

ReasoningCodingAgenticKnowledge
Reasoning
80.6#10 of 16 · 2 tests
Coding
78.3#9 of 13 · 2 tests
Agentic
89.31 test · not ranked
Knowledge
—no data

Model facts

Lab
Z.ai (Zhipu)
Type
Open weights
Released
2026-08-14
Parameters
753B total, 40B active (MoE)
Context
1M
License
GLM-5.3 License (custom; arena.ai lists MIT)

GLM 5.3 hardware requirements and cost

Even at 4-bit it needs about 520 GB (4× H200 (564 GB)), well beyond a home workstation, so most people run it on rented multi-GPU servers. Full BF16 precision needs about 1.7 TB (Multiple nodes (over 1.1 TB)).

Estimated GPU memory for GLM 5.3
PrecisionWeightsWith headroomSmallest setup
BF16 (full precision)1.5 TB1.7 TBMultiple nodes (over 1.1 TB)
FP8 / 8-bit753 GB866 GB8× H200 (1,128 GB)
4-bit (GGUF Q4_K_M)452 GB520 GB4× H200 (564 GB)

Estimate: parameters × bytes per parameter, plus 15% for the KV cache and runtime at modest context lengths. Long contexts and large batches need more.

Lighter or stronger open models

Open models ranked just above and below on the composite index, with their 4-bit memory needs.

GLM 5.3 speed and pricing

Measured figures link to their source; estimates are labelled.

API output speed
88 tok/sArtificial Analysis median · Artificial Analysis
Time to first token
1.09 sOpenRouter model page provider stats
API price per 1M tokens
$1.40 in · $4.40 outlab list price · Z.ai pricing docs
Compare speed and price across all models

Score by test

Rank is among every model with a published score on that test.

GPQA Diamond91.7%GLM 5.3 (max); competitor comparison column#9 of 19 · 4.4 behind GPT-6 AstraHumanity's Last Exam42.3%no tools, text-only subset (2,158 q); competitor comparison column#15 of 23 · 22.1 behind Claude Opus 5.5Humanity's Last Exam (with tools)62.5%with tools#6 of 17 · 7.5 behind Kimi K3.1 ProTerminal-Bench88.2%Terminal-Bench 2.1#4 of 13 · 2.4 behind DeepSeek V4.1 FlashTerminal-Bench 4.041.9%Terminal-Bench 4, Max, mini-swe-agent harness#8 of 16 · 28.7 behind Claude Sonnet 5.5LMArena Text Elo1481GLM-5.3 (Max), ±5, board dated 2026-10-27#6 of 21 · 39 behind Gemini 4 Argon

Nearby in the composite ranking

Models with the closest composite index.

  1. 1GPT-6.1 Sol86.5
  2. 2Qwen3.6 27B85.8
  3. 3Qwen3.8 27B83.2
  4. 4GLM 5.382.7
  5. 5DeepSeek V4.1 Flash79.8
  6. 6Gemma 4 31B79.2
  7. 7Step 5 Preview78.7

Head-to-head comparisons

Side-by-side pages built from the tests both models report.

Sources

ModelTestScoreSettingSource
GLM 5.3GPQA Diamond91.7GLM 5.3 (max); competitor comparison columnMeta Muse Spark Open model card (comparison column)
GLM 5.3Humanity's Last Exam42.3no tools, text-only subset (2,158 q); competitor comparison columnMiniMax M3.1 Flash model card (comparison column)
GLM 5.3Humanity's Last Exam (with tools)62.5with toolsHugging Face model card (GLM-5.3)
GLM 5.3Terminal-Bench88.2Terminal-Bench 2.1Hugging Face model card (GLM-5.3)
GLM 5.3Terminal-Bench 4.041.9Terminal-Bench 4, Max, mini-swe-agent harnessHugging Face model card (GLM-5.3), evaluation results
GLM 5.3LMArena Text Elo1481GLM-5.3 (Max), ±5, board dated 2026-10-27LMArena (arena.ai) text leaderboard

Scores are published results from model cards, lab reports, and public leaderboards, each linked to its source. Settings such as thinking mode or test version differ between sources and are shown next to each score. OpenLLM Buddy catalog models are uncensored fine-tunes; their scores are those of the official base model. Data collected 2026-10-30. How the index works