GLM 4.5 — Benchmarks

Benchmark scores for GLM 4.5 aggregated from public leaderboards (8 boards), with how it ranks among open models and how it compares with its nearest neighbours. See hardware requirements for what you need to run it.

llmrun Composite Score →

100.9

90% range 96.5–105.2 · from 7 boards

Coding98.790% range 89–109
Agents105.490% range 99–112
MathNot enough data yet
ScienceNot enough data yet
ReasoningNot enough data yet

GLM 4.5 reasoning benchmarks

Multi-step logic and problem-solving — can the model think through a hard problem rather than recall an answer.

Fiction.LiveBench
Fiction.LiveBench is a long-context comprehension test: models answer questions that require tracking characters, events and relationships across a long story. llmrun shows the 16k-token score, the column Epoch AI reports as its headline.
58.3%
58.3%
median 61.1%open leader 86.1%
Overall best: GPT 5 (Aug 07, 2025, medium) 97.2%
46th percentile · #14 of 25 open · #37 of 58 all
#14 of 25
#37 of 58
Independent
Epoch AI · Jul 2025

GLM 4.5 coding benchmarks

Writing, editing, and fixing real code — measured by whether the resulting program actually runs and passes tests.

SWE-bench Verified
SWE-bench Verified tests whether a model can resolve real GitHub issues from popular open-source Python projects, scored on the official swebench.com leaderboard as the percentage of human-validated issues actually fixed. It is the headline measure of practical, agentic software-engineering ability — where open-weight models like Qwen3-Coder, GLM-4.6, Kimi K2 and DeepSWE are now competitive with the frontier.
64.2%
64.2%
median 62.2%open leader 75.8%
Overall best: live-SWE-agent + Claude 4.5 Opus medium (20251101) 79.2%
58th percentile · #9 of 20 open · #65 of 162 all
#9 of 20
#65 of 162
Independent
SWE-bench · Jul 2025
SWE-bench Bash Only
SWE-bench Bash Only runs the SWE-bench Verified issues through a minimal, single-tool bash agent — no specialised scaffolding — so the score reflects the model's own agentic coding ability rather than the harness around it. A cleaner, harder read on raw software-engineering skill.
54.2%
54.2%
median 55.4%open leader 63.4%
Overall best: Claude 4.5 Opus (high reasoning) 76.8%
38th percentile · #6 of 9 open · #33 of 48 all
#6 of 9
#33 of 48
Independent
SWE-bench · Aug 2025
WeirdML
WeirdML asks models to write machine-learning code for small, deliberately unusual datasets, then runs the code and scores the accuracy it actually achieves. Built by Håvard Tveit Ihle, it rewards practical problem-solving over recognizing a familiar task.
40.6%
40.6%
median 39.5%open leader 82.6%
Overall best: GPT 6 Astra Promax 93.6%
52nd percentile · #23 of 47 open · #106 of 162 all
#23 of 47
#106 of 162
Independent
Epoch AI · Aug 2025
ALE-Bench
ALE-Bench, from Sakana AI and AtCoder, tests long-horizon algorithmic optimisation using problems drawn from AtCoder Heuristic Contests, where there is no single correct answer and solutions are scored on quality. Results are reported as an AtCoder-style performance rating, where higher is better.
344.8
344.8
median 754.6open leader 1932.8
Overall best: GPT 6 Astra Max 2951.3
13th percentile · #42 of 48 open · #116 of 129 all
#42 of 48
#116 of 129
Lab-reported
Epoch AI · Aug 2025

GLM 4.5 knowledge benchmarks

Breadth and depth of factual knowledge across many subjects, paired with the reasoning to apply it.

Humanity's Last Exam
Humanity's Last Exam (HLE) is a set of extremely difficult, expert-written questions across many fields, designed so that even frontier models score low. It is built to stay hard as models improve, measuring the true knowledge frontier.
8.3%
8.3%
median 8.2%open leader 24.4%
Overall best: GPT 6 Astra (unspecified) 54.8%
67th percentile · #2 of 4 open · #35 of 51 all
#2 of 4
#35 of 51
Lab-reported
Epoch AI · Aug 2025
MMLU-Pro
MMLU-Pro is a harder, cleaned-up successor to MMLU with ten answer choices and more reasoning-heavy questions across 14 subjects, measuring broad knowledge and reasoning together.
84.6%
84.6%
median 52.0%open leader 88.0%
Overall best: Gemini-3.1-Pro 91.2%
94th percentile · #10 of 146 open · #36 of 257 all
#10 of 146
#36 of 257
Independent
TIGER-Lab

GLM 4.5 instruction following benchmarks

How precisely a model obeys explicit, verifiable instructions — format, length, and content constraints — and, more broadly, how much people prefer its responses head-to-head in blind human voting.

LMArena Text
Arena (arena.ai, formerly LMArena / Chatbot Arena) ranks models by millions of blind, head-to-head human votes on real prompts. The Arena Score shown here is a Bradley-Terry rating computed with Arena's default style control, so higher means people preferred that model's answers more often — it's a relative rating, not a percentage, so compare it only to other models on this board.
1410.9
1410.9
median 1321.4open leader 1488.2
Overall best: Gemini 4 Carbon (High) 1527.5
77th percentile · #44 of 188 open · #146 of 408 all
#44 of 188
#146 of 408
Arena
Arena · Oct 2026

Score vs VRAM in the ≤256 GB tier

GLM 4.5
Other models
Best Score for its VRAM
x: VRAM (GB, log scale) · y: llmrun Score

#38 of 89 that fit ≤256 GB

Each dot is a model at the best quant that fits ≤256 GB. The line steps up to the best Score available at each amount of VRAM; a dot below it is beaten by a model that needs no more memory.

Full ≤256 GB ranking →

GLM 4.5 vs its neighbours

Same boards for the nearest open models per llmrun Score. The highest score in each row is marked.

Board	GLM 4.5this model	Qwen3.8 27Bbest in classQ4_K_M · 18 GB	DeepSeek V3.1nearest by ScoreQ4_K_M · 412 GB	Qwen3 235B A22Bnearest by ScoreQ4_K_M · 143 GB	DeepSeek R1nearest by ScoreQ4_K_M · 412 GB
llmrun Score	100.9	122.0	100.2	100.0	98.9
Fiction.LiveBench	58.3%	—	52.8%	67.7%	69.4%
WeirdML	40.6%	—	38.4%	37.3%	36.5%
MMLU-Pro	84.6%	—	84.8%	68.2%	84.0%
LMArena Text	1410.9	1435.6	1416.9	1402.1	1398.1
Compare head to head		Compare	Compare	Compare	Compare

Scores aggregated from public benchmark sources (Epoch AI: 4 boards, SWE-bench: 2 boards, Arena: 1 board, TIGER-Lab: 1 board) — each linked from the benchmark pages. llmrun does not run these benchmarks. Percentiles and ranks are among open models in the llmrun catalogue; the median and leader marks refer to the same field.