VerdictPal
Join

Models / Zhipu

GLM-5.3

GLM · Released 2026-08-14

Z.ai's August 2026 coding refresh of the GLM-5.2 base. Artificial Analysis Intelligence Index 59.5 on the max-effort endpoint. API list price $1.40 / $4.40 per 1M tokens — same slot as GLM-5.2.

Current Z.ai coding-plan and API model for long-horizon agent work. Prefer sourced ledger rows over vendor coding/cyber tables; AA scores sync from the public snapshot. Open weights were promised after a safety delay, not at launch.

#4 of 166 on AA Index · current snapshot
60 AA Index · 11 benchmark rows · 2 sourcesPosition is benchmark-specific — not a cross-family or cross-source ranking.
Composite54
ContextUnlisted
Input / 1M$1.4
Output / 1M$4.4
Knowledge cutoffunpublished
Output speed73 t/s
TTFT (default API)1.56s
Statussolid

Family profile

Best published score in each covered benchmark family.

77/100 avg
Reasoning
92
Coding
57
Agentic
84
Long context
76

4 tested benchmark families

Editor's note
Recorded from the 2026-08-24 AA snapshot (slug glm-5-3, max effort, index 59.5) plus the Z.ai launch post. Same base as GLM-5.2; gains are post-training. Vendor Terminal Bench / Z.ai Code Bench / cyber figures stay off the ledger until they appear in a mapped source. Editorial depth still needs a desk pass.
Benchmark placements

Where GLM-5.3 places on each public benchmark source that publishes a row. Rank counts every model with a latest row in the same test — not a universal quality score.

BenchmarkFamilyRankScoreSourceDate
τ³-Bench BankingAgentic#3/ 10650Artificial Analysis2026-09-01
Terminal-Bench-ScienceAgentic#6/ 98Terminal-Bench-Science 0.1 announcement2026-08-27
SciCodeCoding#7/ 18057Artificial Analysis2026-09-01
Artificial Analysis Intelligence IndexReasoning#9/ 18460Artificial Analysis2026-09-01
CritPtReasoning#11/ 1219Artificial Analysis — CritPt2026-09-01
Terminal-BenchAgentic#14/ 18484Artificial Analysis2026-09-01
GPQA DiamondReasoning#25/ 18592Artificial Analysis2026-09-01
Humanity's Last ExamReasoning#26/ 18042Artificial Analysis2026-09-01
AA-LCRLong context#38/ 17976Artificial Analysis2026-09-01
AA time to first tokenPerformance#41/ 185Artificial Analysis2026-09-01
AA output speedPerformance#54/ 185Artificial Analysis2026-09-01
Family context

The three highest-scoring models with pages in each capability family. Where GLM-5.3 shows up, it's highlighted.

ReasoningGLM-5.3 · 92
  1. 1
    Claude Mythos Preview
    95
  2. 2
    Gemini 3.7 Flash
    95
  3. 3
    Grok 4.6
    95
CodingGLM-5.3 · 57
  1. 1
    GPT 5.5
    96
  2. 2
    Claude Opus 4.8
    96
  3. 3
    Claude Opus 4.6
    96
AgenticGLM-5.3 · 84
  1. 1
    GPT 5.5
    100
  2. 2
    Claude Opus 4.8
    100
  3. 3
    Claude Fable 5
    100
Long contextGLM-5.3 · 76
  1. 1
    Muse Spark 1.2
    83
  2. 2
    Kimi K3
    83
  3. 3
    Gemini 3.5 Flash
    81
Newest receipts

Benchmark rows added to the public ledger for GLM-5.3 in the last 120 days. Older rows live in the full table below.

11 recent
BenchmarkFamilyScoreSourceDays ago
CritPtReasoning19.1%Artificial Analysis — CritPt2
AA-LCRLong context76.33%Artificial Analysis2
AA output speedPerformance73 t/sArtificial Analysis2
AA time to first tokenPerformance1.56sArtificial Analysis2
Artificial Analysis Intelligence IndexReasoning59.5%Artificial Analysis2
GPQA DiamondReasoning91.7%Artificial Analysis2
Humanity's Last ExamReasoning42.3%Artificial Analysis2
SciCodeCoding56.5%Artificial Analysis2
τ³-Bench BankingAgentic50.31%Artificial Analysis2
Terminal-BenchAgentic83.9%Artificial Analysis2
Terminal-Bench-ScienceAgentic8.1%Terminal-Bench-Science 0.1 announcement7
Family coverage

A high composite that hides a weak family is a trap. These bars surface the families where this model hasn't been publicly tested, and where it leads.

Reasoning92
Coding57
Agentic84
Long context76
Price vs. performance
0255075100$0.10$0.30$1$3$10$30input USD / 1M tokens · log scale →← intelligence

Hover or tab any dot for name, score, and input price.

AnthropicOpenAIMistralAlibabaDeepSeekGoogleMoonshotMetaMiniMaxxAIZhipuXiaomiOtherMeituanCohereNVIDIA

Full benchmark ledger

Every catalog benchmark for GLM-5.3. Scores link to the original source; gaps mean no public row exists yet.

11 sourced rows

11 of 37 catalog benchmarks have a sourced row for GLM-5.3.

Reasoning

Coding

Agentic

Long context

Performance

Changelog
  1. GLM-5.3 announced on the GLM Coding Plan. Open weights delayed for safety review.
  2. Hosted API listed on Artificial Analysis (max-effort index 59.5, $1.40 / $4.40 per 1M tokens).
Sources