Skip to main content

Model profile

GLM-5

Z.AISupersededReleased Mar 1, 2026
Data verified
Superseded:Z.AI has released newer models in this line —GLM-5.1
Overall Score
66.06Public #28 of 200Verified #24 of 99
Arena Elo
1457
Eligible category ranks
6of 8
Price (1M tokens)
$1 in / $3.2 out
API pricing
Speed
74tok/s
Context
200K

Evidence coverage

49 of 321 tracked benchmarks are published. 8 are verified and 41 provisional. 8 of 8 categories are measured.

Updated July 20, 2026Methodology
Published / tracked
49 / 321
Verified
8
Provisional
41
Categories with evidence
8 / 8

Evidence by category

  • Agentic13 benchmarks
    Mixed evidence
  • Coding7 benchmarks
    Mixed evidence
  • Reasoning4 benchmarks
    Reported
  • Knowledge12 benchmarks
    Reported
  • Math8 benchmarks
    Mixed evidence
  • Multilingual2 benchmarks
    Reported
  • Multimodal1 benchmark
    Reported
  • Inst. Following2 benchmarks
    Reported
Open WeightSelf-hostNon-Reasoning
Confidence:
Very high
base

GLM-5 ranks #28 out of 200 models on the public leaderboard with an overall score of 66.06/100. It also ranks #24 out of 99 on the verified leaderboard. This places it in the mid-tier of AI models, with strengths in specific benchmark categories.

GLM-5 is a open weight model with a 200K token context window. It processes queries without explicit chain-of-thought reasoning, offering faster response times and lower token usage.

GLM-5 sits inside the GLM-5 family alongside GLM-5.2, GLM-5.1, GLM-5 (Reasoning), GLM-5-Turbo, GLM-5V-Turbo. This profile currently has 49 of 321 tracked benchmarks. BenchLM only exposes non-generated benchmark rows publicly, so missing categories stay blank until a sourced evaluation is available.

Its strongest category is Multilingual (#6), while its weakest is Coding (#24). This performance profile makes it a well-rounded choice across a range of tasks.

Peer position

Exact provisional scores and ranks for the closest listed peers. A score can appear before a model clears the evidence threshold for a rank, so equal scores can have different rank states.

Range 65.266.89

  1. GLM-5-Turbo
    Z.AI
    #2466.89
    GLM-5-Turbo is #24 with a score of 66.89.
    Compare
  2. GPT-5.4 nano
    OpenAI
    #2566.79
    GPT-5.4 nano is #25 with a score of 66.79.
    Compare
  3. GPT-5.3 Codex
    OpenAI
    #2666.69
    GPT-5.3 Codex is #26 with a score of 66.69.
    Compare
  4. Claude Opus 4.7 (Adaptive)
    Anthropic
    #2766.27
    Claude Opus 4.7 (Adaptive) is #27 with a score of 66.27.
    Compare
  5. GLM-5Current model
    Z.AI
    #2866.06
    GLM-5 is #28 with a score of 66.06.
  6. Claude Sonnet 5
    Anthropic
    #2965.32
    Claude Sonnet 5 is #29 with a score of 65.32.
    Compare
  7. Qwen3.6 Plus
    Alibaba
    #3065.2
    Qwen3.6 Plus is #30 with a score of 65.2.
    Compare

Category percentile

More

Relative position among models eligible for each sourced category. A higher percentile means a stronger position within that category's ranked cohort; 100 is highest.

  1. Multilingual58%
    Eligible cohort rank #6 of 13Category score 48.7
  2. Math0%
    Eligible cohort rank #7 of 7Category score 57.2
  3. Inst. Following68%
    Eligible cohort rank #13 of 38Category score 86.5
  4. Knowledge75%
    Eligible cohort rank #14 of 52Category score 80.8
  5. Agentic81%
    Eligible cohort rank #23 of 119Category score 54.8
  6. Coding81%
    Eligible cohort rank #24 of 122Category score 59.4

Category evidence

Scores and ranks appear only where this model has published benchmark evidence. Categories without displayable source records remain not measured.

Category scores, ranks, weighting, benchmark coverage, and evidence status
CategoryScore
AgenticRank #23 of 119Percentile 81stWeight 22%13 benchmarksMixed sources54.8
CodingRank #24 of 122Percentile 81stWeight 20%7 benchmarksMixed sources59.4
ReasoningRank Not rankedWeight 17%4 benchmarksReported78.9
KnowledgeRank #14 of 52Percentile 75thWeight 12%12 benchmarksReported80.8
MathRank #7 of 7Percentile 0thWeight 5%8 benchmarksMixed sources57.2
MultilingualRank #6 of 13Percentile 58thWeight 7%2 benchmarksReported48.7
MultimodalRank Not rankedWeight 12%1 benchmarkReported68.8
Inst. FollowingRank #13 of 38Percentile 68thWeight 5%2 benchmarksReported86.5

Chatbot Arena performance

Scroll horizontally to inspect confidence intervals and vote counts.

Chatbot Arena Elo, confidence interval, and vote count by evaluation view
ViewEloConfidence intervalVotes
Text Overall1457±4.427,814
Coding1498±7.57,210
Math1443±14.51,603
Instruction Following1447±6.88,910
Creative Writing1445±9.34,519
Multi-turn1473±9.24,437
Hard Prompts1478±5.317,146
Hard Prompts (English)1487±7.08,430
Longer Query1470±6.610,192

Benchmark Details

Rows below have a displayable published verification record. Each source link and provenance note remains in the page HTML while its category is closed. Source-unverified manual rows and generated rows stay hidden.

Agentic13 benchmarks
Terminal-Bench 2.0Provider exact
56.2%Weighted 38%
Source: Z.AI: GLM-5Provenance: Provider exact
Claw-EvalSecondary exact
57.7%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
QwenClawBenchSecondary exact
54.1%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
τ³-bench resultsSecondary exact

τ³-Bench Tool-Agent-User Evaluation

65.6%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
DeepPlanningSecondary exact
14.6%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
ToolathlonSecondary exact
38%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
MCP AtlasSecondary exact
31.1%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
MCP-TasksSecondary exact
60.8%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
WideResearchSecondary exact
69.8%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
τ²-bench resultsReported

τ²-Bench Tool-Agent-User Evaluation

98.2%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
CyberGymBenchmark exact
43.2%Display only
Source: CyberGym leaderboardProvenance: CyberGym Level 1 reports Claude Code with GLM-5 on the public leaderboard. BenchLM stores the reported target-vulnerability reproduction success rate on the local cyberGym key.
APEX-Agents-AAReported
14.5%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Gert LabsBenchmark exact

Gert Labs Composite Game Benchmark

50.99%Display only
Source: Gert Labs rankingsProvenance: Gert Labs reports this composite leaderboard score in the public rankings API. BenchLM scales the source gscore from 0-1 to 0-100 and stores it as a display-only agentic benchmark.
Coding7 benchmarks
SWE-RebenchBenchmark exact
62.8%Weighted 20%
Source: SWE-Rebench leaderboardProvenance: Public SWE-Rebench leaderboard lists GLM-5 at 62.8% resolved rate.
SWE-bench VerifiedProvider exact

Software Engineering Benchmark Verified

77.8%Weighted 16%
Source: Z.AI: GLM-5Provenance: Provider exact
SWE-bench ProSecondary exact
55.1%Weighted 10%
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
SWE-bench Verified*Secondary exact

SWE-bench Verified (mini-swe-agent-v2)

72.8%Display only
Source: Arcee Trinity-Large-Thinking comparison tableProvenance: Secondary exact
SWE MultilingualSecondary exact
73.3%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
React Native EvalsBenchmark exact
74.8%Display only
Source: React Native Evals leaderboardProvenance: React Native Evals reports this exact overall score for GLM 5 in the public dashboard run finished on 2026-04-28.
AA-SciCodeReported

Artificial Analysis SciCode

46.2%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Reasoning4 benchmarks
LongBench v2Secondary exact
60.8%Weighted 38%
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
AI-NeedleSecondary exact
63.3%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
AA-LCRReported

Artificial Analysis Long Context Reasoning

63.3%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
CritPtReported

Critical Physics Tasks

2.0%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Knowledge12 benchmarks
HLESecondary exact

Humanity's Last Exam

50.4%Weighted 45%
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
MMLU-ProSecondary exact

Massive Multitask Language Understanding Professional

85.7%Weighted 30%
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
GPQASecondary exact

Graduate-Level Google-Proof Q&A

86%Weighted 7%
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
SuperGPQASecondary exact

SuperGPQA: Scaling LLM Evaluation Across 285 Graduate Disciplines

66.8%Weighted 7%
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
GPQA-DSecondary exact

GPQA Diamond

86.0%Display only
Source: Arcee Trinity-Large-Thinking comparison tableProvenance: Secondary exact
MMLU-Pro (Arcee)Secondary exact

MMLU-Pro first-party comparison snapshot

85.8%Display only
Source: Arcee Trinity-Large-Thinking comparison tableProvenance: Secondary exact
Artificial Analysis Intelligence IndexReported
39.5%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-GPQA DiamondReported

Artificial Analysis GPQA Diamond

82.0%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-HLEReported

Artificial Analysis Humanity's Last Exam

27.2%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience IndexReported

Artificial Analysis Omniscience Index

2.0%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience AccuracyReported

Artificial Analysis Omniscience Accuracy

26.9%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience Hallucination RateReported

Artificial Analysis Omniscience Hallucination Rate

34.0%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Math8 benchmarks
FrontierMath v2 (Tiers 1-3)Benchmark exact

FrontierMath v2 Tiers 1-3

16.434%Weighted 30%
Source: Epoch AI FrontierMath v2 leaderboardProvenance: Epoch AI reports FrontierMath v2 Tiers 1-3 at 16.434% for glm-5. BenchLM selects the highest published thinking effort for the model and stores the v2 benchmark slice separately.
AIME26Secondary exact

AIME 2026

95.8%Weighted 25%
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
HMMT Feb 2026Secondary exact

Harvard-MIT Mathematics Tournament February 2026

86.4%Weighted 25%
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
FrontierMath v2 (Tier 4)Benchmark exact

FrontierMath v2 Tier 4

2.100%Weighted 10%
Source: Epoch AI FrontierMath v2 leaderboardProvenance: Epoch AI reports FrontierMath v2 Tier 4 at 2.1% for glm-5. BenchLM selects the highest published thinking effort for the model and stores the v2 benchmark slice separately.
AIME25 (Arcee)Secondary exact

AIME25 first-party comparison snapshot

93.3%Display only
Source: Arcee Trinity-Large-Thinking comparison tableProvenance: Secondary exact
HMMT Feb 2025Secondary exact

Harvard-MIT Mathematics Tournament February 2025

97.5%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
HMMT Nov 2025Secondary exact

Harvard-MIT Mathematics Tournament November 2025

96.9%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
MMAnswerBenchSecondary exact
82.5%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
Multilingual2 benchmarks
MMLU-ProXSecondary exact
83.1%Weighted 100%
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
NOVA-63Secondary exact
55.1%Display only
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
Multimodal1 benchmark
Design Arena WebsiteReported

Design Arena Website Elo

1280Display only
Source: OpenRouter model benchmarksProvenance: Display-only Design Arena Website Elo synced from OpenRouter model benchmark metadata. It is excluded from BenchLM weighted scoring.
Inst. Following2 benchmarks
IFEvalSecondary exact

Instruction-Following Eval

92.6%Weighted 35%
Source: Qwen3.6-Plus comparison tableProvenance: Secondary exact
AA-IFBenchReported

Artificial Analysis IFBench

72.3%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.

Frequently Asked Questions

How does GLM-5 perform overall in AI benchmarks?

GLM-5 currently ranks #28 out of 200 models on BenchLM's provisional leaderboard with an overall score of 66.06. It also ranks #24 out of 99 on the verified leaderboard. It is created by Z.AI. Its published context window is 200K.

Is GLM-5 good for knowledge and understanding?

GLM-5 ranks #14 out of 52 models in knowledge and understanding benchmarks with an average score of 80.8. There are stronger options in this category.

Is GLM-5 good for coding and programming?

GLM-5 ranks #24 out of 122 models in coding and programming benchmarks with an average score of 59.4. There are stronger options in this category.

Is GLM-5 good for mathematics?

GLM-5 ranks #7 out of 7 models in mathematics benchmarks with an average score of 57.2. It is among the top performers in this category.

Is GLM-5 good for reasoning and logic?

GLM-5 has visible benchmark coverage in reasoning and logic, but BenchLM does not currently assign it a global category rank there.

Is GLM-5 good for agentic tool use and computer tasks?

GLM-5 ranks #23 out of 119 models in agentic tool use and computer tasks benchmarks with an average score of 54.8. There are stronger options in this category.

Is GLM-5 good for multimodal and grounded tasks?

GLM-5 has visible benchmark coverage in multimodal and grounded tasks, but BenchLM does not currently assign it a global category rank there.

Is GLM-5 good for instruction following?

GLM-5 ranks #13 out of 38 models in instruction following benchmarks with an average score of 86.5. There are stronger options in this category.

Is GLM-5 good for multilingual tasks?

GLM-5 ranks #6 out of 13 models in multilingual tasks benchmarks with an average score of 48.7. It is among the top performers in this category.

Is GLM-5 open source?

Yes, GLM-5 is an open weight model created by Z.AI, meaning it can be downloaded and run locally or fine-tuned for specific use cases.

Which sibling models are related to GLM-5?

GLM-5 belongs to the GLM-5 family. Related variants on BenchLM include GLM-5.2, GLM-5.1, GLM-5 (Reasoning), GLM-5-Turbo, GLM-5V-Turbo.

Does GLM-5 have full benchmark coverage on BenchLM?

Not yet. GLM-5 currently has 49 published benchmark scores out of the 321 benchmarks BenchLM tracks. BenchLM only exposes non-generated public benchmark rows, so missing categories stay blank until a sourced evaluation is available.

What is the context window size of GLM-5?

GLM-5 has a published context window of 200K, which determines how much text it can process in a single interaction.

Last updated: July 20, 2026 · Runtime metrics stay blank until BenchLM has a sourced snapshot.

Choose with this week’s evidence

Join 2,000+ readers for ranking moves, new releases, pricing changes, and the evidence behind them.

Free. One email per week.