Zhipu AIChina· released Aug 2026· open weights

GLM 5.3

Name: GLM 5.3
Author: Zhipu AI

Provisional: 12 comparable tests so far. Ranked once it reaches 6 across 3 counting categories.

Zhipu AI, China. One of 7 models from this lab on the Index.

Third-party publishedLow confidence (thin coverage)

Available in Theo as Theo Open Sage

Theo orchestrates it for the steps it does best and always shows which engine answered.

Use it in Theo

Snapshot September 8, 2026

OpenCharts Index

60.9

Provisional: not enough comparable tests or counting categories to rank yet · 12 comparable tests · 1 counting categories

Drop any single test and the Index lands between 50.8 and 71.2.

Needs more results (2 of 8 so far)

Averaging 51.7 on the 2 it has; the score opens at 4 of 8.

Distance to today's frontier on the hardest tests. Not a measure of intelligence.

What it's known for

GLM 5.3, in 12 results.

Strongest in math (60.9). Its widest gap is on FrontierMath Tier 4, 70.0% behind the best published result. No comparable results yet in multimodal and long context. Between 35% and 40% behind the best published results, on average.

Strongest category

Math

Competition and research mathematics, graded on the final answer or the proof.

60.9

Mock AIME 2024–202591.1% · 91% of best
FrontierMath (Tiers 1–3)68.8% · 73% of best
ProofBench49.0% · 49% of best
FrontierMath Tier 429.3% · 30% of best

Mean of 4 of the category's 5 comparable tests. Counts toward the Index.

Best single result

GPQA Diamond

Answer a PhD-level multiple-choice question in physics, chemistry or biology that a web search will not settle.

90.9%

93.1% of the best (95.8%), above chance · #22 of 78.

Expert-level science questions are the closest thing to asking a specialist colleague. A model that gets them right can check a technical claim instead of echoing it.

Where it trails

FrontierMath Tier 4

Solve the hardest FrontierMath tier: problems that take expert mathematicians days.

29.3%

70.0% behind the best published result (97.6%) · #22 of 51.

The deepest end of the set. Progress here signals genuinely new capability.

Not measured yet

No comparable result yet in these 2 categories. A blank is a blank, never a zero, and it does not lower the Index.

Multimodal
Long context

Covered, but not averaged into the Index:

Reasoning2 of 6 tests · needs 3 · mean 76.2
Knowledgesingle-test category, shown beside the Index · 54.2
Coding3 of 7 tests · needs 4 · mean 74.3
Agentic1 of 8 tests · needs 4 · mean 89.9
Human preferencesingle-test category, shown beside the Index · 92.8

Every score, with its source

See it at work

What GLM 5.3 was asked to do, and how it did.

Every test it has taken, by category, with the kind of task it faced, who produced the number and where the result landed against the best published one. The examples are original and illustrative, never items from the datasets themselves.

Human preference · mean 92.8 across 1 comparable test · a single-test category, shown beside the Index and never averaged into it

Text ArenaLeaderboard

Answer the same prompt as a rival model; real users vote blind on which answer they prefer.

1,482

46% expected win rate against the board leader (1,507)
#16 of 92

max · board 2026-09-02

Source: Arena

Every result is read against the best published one on its test: percent scores as a share of the best above chance, Arena ratings as the expected win rate against the board leader, open-ended values on a log scale. Sitting a harder exam never lowers a model. Every number links to the publisher that produced it.

Head to head

GLM 5.3 against whoever you pick.

Choose a rival. Every counted test both have taken appears side by side as a share of the best published result on that test, with the category means above. A win is a higher share; a gap under 2 points is a tie, because one item on a 45-problem exam is 2.2 points and publishers' own standard errors are of that order.

GLM 5.3

provisional·Index 60.9C

0 : 9

wins · 1 tie · 10 shared

GPT-6 Astra

#1·Index 99.0A+

GPT-6 Astra comes out ahead on 9 of the 10 tests both have taken.

Category means

Reasoning
76.2
99.6
Knowledge
54.2
100.0
Coding
74.3
97.7
Math
60.9
99.8
Agentic
89.9
98.5
Human preference
92.8
—

Shared tests · biggest gaps first

FrontierMath Tier 4Math
29.3%
97.6%
−70.0
ProofBenchMath
49.0%
99.0%
−50.0
Code Arena (WebDev)Coding
1,609
1,797
−49.4
SimpleQA VerifiedKnowledge
41.0%
75.6%
−45.8
CritPtReasoning
19.1%
31.7%
−38.9
FrontierMath (Tiers 1–3)Math
68.8%
93.7%
−26.6
WeirdMLCoding
75.4%
92.9%
−18.8
Mock AIME 2024–2025Math
91.1%
100.0%
−8.9
GPQA DiamondReasoning
90.9%
95.8%
−6.9
SciCodeCoding
56.5%
56.5%
tie

Bars are distances to the best published result on each test (percent scores above chance, Arena ratings as win rate against the leader, open-ended values on a log scale). Raw values as the publishers report them.

Open GPT-6 Astra

Scores

Every test, every source

The best published run per benchmark, the raw value as the publisher reports it, who produced the number, how close it comes to the best published result on that test, and where the model sits among every model scored on it. Percent scores are read above chance, Arena ratings as a win rate against the board leader, open-ended values on a log scale. A test with too few models to compare against is shown but not counted.

Category means

Reasoning76.2 · 2/6
Knowledgebeside54.2 · 1/1
Coding74.3 · 3/7
Math60.9 · 4/5
Agentic89.9 · 1/8
Multimodal—
Human preferencebeside92.8 · 1/1
Long context—

Mean · tests taken of the category's comparable tests. A faded category either rests on fewer than half of them or is a single-test category (marked “beside”), which is shown next to the Index and never averaged into it.

GPQA DiamondEpoch-runReasoning

90.9%

93.1% of the best (95.8%), above chance
70th percentile of 78 · n = 198

max

Source: Epoch AI — AI Benchmarking Hub

CritPtLeaderboardReasoning

19.1%

59.3% of the best (32.3%)
78th percentile of 78

max

Source: Epoch AI — AI Benchmarking Hub

SimpleQA VerifiedEpoch-runKnowledge

41.0%