Inference
EN
Z.AI

GLM-5

GLM-5 is a standard model from Z.AI. 38 benchmarks count toward its index, in 7 categories.

  • Open weights
  • Standard
  • Input: text
  • Released 2026-02
RankedEnough results for an index.
IndexAll categories, out of 100.51.6±3.6Median 51.0
CoverageIndex weight that has results.90%
SpeedOutput tokens per second.45/sMedian 67/s
First tokenSeconds to the first output token.1.5 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$1.00Median $0.60
Output / 1MUS dollars per 1M output tokens.$3.20Median $2.55
ContextMaximum tokens in one request.205KMedian 262K
EloLMArena rating and rank.1446±4 #61Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
z-ai/glm-5
Max outputMaximum output tokens per request.
131K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.20
WeightsLink to the published weights.
Hugging Face (opens in a new tab) ↗
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
11 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
VultrCloud$0.40$1.75203K
GMI CloudHost$0.60$1.92203K
Deep InfraDeprecatedHost$0.60$2.08203K
SiliconFlowHost$0.95$2.55205K
Zhipu AIMaker$0.598$2.69205K
Amazon BedrockCloud$1.00$3.20203K
Google Vertex AIDeprecatedCloud$1.00$3.20200K
Novita AIHost$1.00$3.20203K
OpenRouterRouter$1.00$3.20205K
Together AIDeprecatedHost$1.00$3.20203K
Z.AIMaker$1.00$3.20205K

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Run locally

753.86B parameters
16-bit
1584.1 GB
8-bit
842.0 GB
4-bit
480.9 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy needs more than 128 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.14 counted51.6119 / 218
CodingCode writing and repair.4 counted53.5110 / 209
ReasoningLogic problems and puzzles.5 counted48.3129 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.5 counted51.8122 / 268
MultilingualTasks in many languages.1 counted52.5—
InstructionTasks with strict prompt rules.2 counted53.130 / 63
MathMath problems.7 counted52.576 / 168

Results

38 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
τ²-Bench Tool-Agent-User EvaluationAgentic98.2%7 / 149
Harvard-MIT Mathematics Tournament February 2025Math97.5%1 / 15
Harvard-MIT Mathematics Tournament November 2025Math96.9%1 / 12
SWE-bench Verified · SWE-bench/experimentsCoding72.8%7 / 65
τ²-bench AirlineAgentic82.5%3 / 22
Massive Multitask Language Understanding ProfessionalKnowledge85.7%16 / 75
Instruction-Following EvalInstruction92.6%16 / 70
AIME 2026Math95.8%8 / 34
Artificial Analysis IFBenchInstruction72.3%37 / 146
Software Engineering Benchmark VerifiedCoding77.8%30 / 114
GPQA diamond · Epoch AIKnowledge87.8% ±2.360 / 217
Graduate-Level Google-Proof Q&AKnowledge86.0%48 / 141
MMLU-ProXMultilingual83.1%6 / 17
Artificial Analysis Humanity's Last ExamKnowledge29.3%84 / 222
Gert Labs Composite Game BenchmarkAgentic51.0%20 / 52
Artificial Analysis Long Context ReasoningReasoning75.7%92 / 217
OTIS Mock AIME 2024-2025Math80.0% ±6.086 / 200
τ³-Bench Tool-Agent-User EvaluationAgentic65.6%14 / 32
Artificial Analysis GPQA DiamondKnowledge82.0%92 / 199
FrontierMath-2025-02-28-PrivateMath16.4% ±2.233 / 69
Critical Physics TasksReasoning2.0%105 / 214
MMLU-Pro first-party comparison snapshotKnowledge85.8%3 / 6
LongBench v2Reasoning60.8%13 / 25
Harvard-MIT Mathematics Tournament February 2026Math86.4%15 / 27
Terminal-Bench 2.0 · BenchLMAgentic56.2%24 / 41
Artificial Analysis Omniscience AccuracyKnowledge26.3%96 / 164
SuperGPQA: Scaling LLM Evaluation Across 285 Graduate DisciplinesKnowledge66.8%13 / 22
SWE-bench Verified (mini-swe-agent-v2)Coding72.8%3 / 5
QwenClawBenchAgentic54.1%6 / 10
Humanity's Last ExamKnowledge30.5%47 / 77
ARC-AGI-2 (semi-private)Reasoning4.9%80 / 131
ARC-AGI-1 (semi-private)Reasoning44.7%79 / 126
FrontierMath-Tier-4-2025-07-01-PrivateMath2.1% ±2.135 / 55
GPQA Diamond · BenchLMKnowledge86.0%55 / 84
SWE-bench ProCoding55.1%55 / 83
τ²-bench TelecomAgentic86.8%12 / 18
React Native EvalsCoding74.8%11 / 16
SWE-Bench verified · Epoch AICoding72.1% ±2.122 / 32
Claw-EvalAgentic57.7%32 / 42
Humanity's Last Exam with toolsAgentic50.4%28 / 36
τ²-bench RetailAgentic73.7%18 / 23
BrowseCompAgentic62.0%50 / 63
APEX-Agents-AAAgentic14.5%24 / 30
WideResearchAgentic69.8%13 / 16
ToolathlonAgentic38.0%18 / 22
NL2RepoCoding35.9%30 / 36
Terminal-Bench HardAgentic43.2%21 / 23
τ²-bench BankingAgentic9.8%27 / 28
MCP AtlasAgentic31.1%51 / 52
CyberGymAgentic43.2%29 / 29
FrontierMath-Tier-4-2025-07-01-PublicMath0.0%3 / 36
SWE-bench Multilingual · SWE-bench/experimentsCoding69.7%4 / 14
SWE-bench Multilingual · swebench.comMultilingual69.7%4 / 12
FrontierMath-2025-02-28-PublicMath2.1% ±2.132 / 64
ARC-AGI-2 (public eval)Reasoning5.4%65 / 122
ARC-AGI-1 (public eval)Reasoning58.6%70 / 116
AIME25 first-party comparison snapshotMath93.3%—
MMAnswerBenchMath82.5%—
AI-NeedleReasoning63.3%—
SWE-RebenchCoding62.8%—
MCP-TasksAgentic60.8%—
NOVA-63Multilingual55.1%—
DeepPlanningAgentic14.6%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYSierra ResearchMITSWE-bench teamNo licence statedARC Prize FoundationTerms restrict reusemodels.devMIT

History

5 releases
Index52.3 → 51.6
23 Oct 202612 Nov 2026
Output / 1M$3.20 → $3.20
23 Oct 202612 Nov 2026

Questions

How much does GLM-5 cost?
$1.00 per 1M input tokens, $3.20 per 1M output tokens.
How good is GLM-5?
Its index is 51.6 out of 100. That is place 139 of 283 ranked models.
What is GLM-5 best at?
Math: 52.5 out of 100, place 76 of 168.
How fast is GLM-5?
45 output tokens per second, 1.5 s to the first token.
What is the context window of GLM-5?
205K tokens.
Is GLM-5 open source?
Yes. Its weights are published.
When was GLM-5 released?
February 2026.

Badge

GLM-5 on Inference360: place 139 of 283, index 51.6
HTML
<a href="https://inference360.ai/models/glm-5"><img src="https://inference360.ai/badges/glm-5.svg" alt="GLM-5 on Inference360: place 139 of 283, index 51.6" height="20"></a>
Markdown
[![GLM-5 on Inference360: place 139 of 283, index 51.6](https://inference360.ai/badges/glm-5.svg)](https://inference360.ai/models/glm-5)

Compare with

o351.8MiMo-V2-Pro51.8Qwen3.5-27B51.3GLM-5-Turbo51.9MiniMax M2.751.2

Lower price

dots3-note Preview63.4 · FreeApodex 1.1 Mini55.7 · FreeNemotron 3 Ultra50.8 · Free

More from Z.AI

All models
GLM-5 (Reasoning)UnrankedGLM-4.7-FlashUnrankedGLM-5-Turbo51.9GLM-5V-Turbo48.1GLM-4.750.0