GLM-5

Released

Feb 11, 2026

Price in / out

—

Blended $/1M

—

Context

200K

Parameters: 744B / 40B active. License: MIT.

Benchmark scoresvs GLM-4.7

Benchmark	Score	Δ vs prior	Rank (current)	Configuration	Source
SWE-bench Verified	77.8%	+4.0	#4 of 9	leaderboard	benchlm.ai ↗
GPQA Diamond	86%	—	#10 of 11	leaderboard	benchlm.ai ↗
Humanity's Last Exam	50.4%	+25.6	#6 of 15	leaderboard	benchlm.ai ↗
AIME 2026	95.8%	—	#1 of 2	leaderboard	benchlm.ai ↗
τ²-bench (Telecom)	98.2%	+2.3	#1 of 10	leaderboard	benchlm.ai ↗
MMLU-Pro	85.7%	+1.4	#3 of 4	leaderboard	benchlm.ai ↗
Artificial Analysis Intelligence Index	39.5	+5.8	#6 of 11	leaderboard	benchlm.ai ↗

GLM versions

Timeline

Version	Released	$ in / out	Terminal-Bench	SWE-bench	GPQA	HLE	AIME	τ²-bench
GLM-4.6Superseded	2025-09-30	—	—	—	—	—	—	76.9%
GLM-4.7Superseded	2025-12-22	—	—	73.8%	—	24.8%	—	95.9%
GLM-5Superseded	2026-02-11	—	—	77.8%	86%	50.4%	95.8%	98.2%
GLM-5.1Superseded	2026-04-15	—	56.9%	—	86.2%	52.3%	95.3%	97.7%
GLM-5.2Superseded	2026-06-17	$1.4 / $4.4	67.8%	—	91.2%	54.7%	99.2%	99.1%
GLM-5.3	2026-08-14	$1.4 / $4.4	71.5%	—	—	—	—	—
GLM-5.3 Flash	2026-08-26	—	62.9%	—	—	—	—	—

Compare

Open the Comparison Lab with GLM-5 and its open-weight peers pre-selected, then toggle any benchmark or model.

Every GLM version over time

Shipping with it?

Benchmarks are a starting point. If you run GLM-5 in production, post what it does well and where it falls over; other founders are deciding right now.

Write about GLM-5