LLMBenchmark.ai

Enter opens the highlighted result. Esc closes.

Claude Haiku 4.5 vs GLM 5.2

Updated October 21, 2026. 69 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace Claude Haiku 4.5 Replace GLM 5.2
Claude Haiku 4.5Anthropic
Coding
49.8
Overall
16.9
Output $/M
$5.00
1
GLM 5.2Z.ai
Coding
57.7
Overall
33.7
Output $/M
$7.00
68
vs boards won of 69 shared

Decision reading

Pick GLM 5.2 for coding: it scores 57.7 on the Vibe Coding Index against 49.8 for Claude Haiku 4.5, though on 1 benchmark against 3.

Pick Claude Haiku 4.5 when output tokens cost 1.4x less, or when a solved task costs $0.56 against $0.86.

Model AClaude Haiku 4.5Anthropic49.8/1003 independent benchmarks$0.56 per solved taskModel BGLM 5.2Z.ai57.7/1001 independent benchmark$0.86 per solved task
Shared results
69
Claude Haiku 4.5 only
25
GLM 5.2 only
28
Like-for-like categories
8 of 9

Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.

Which one for your work

Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.

At a glance

Coding index
Claude Haiku 4.5 49.8 from 3 benchmarks, GLM 5.2 57.7 from 1 benchmark.
Widest category
Agentic work, 31.5 points apart on average over 5 shared benchmarks.
Widest single gap
Vibe Code Bench (Vals AI): Claude Haiku 4.5 11.4% against GLM 5.2 64.0%, 52.6 points apart on a 0 to 100 scale.
Output price
Claude Haiku 4.5 $5.00 against GLM 5.2 $7.00 per million output tokens, 1.4x apart.
Context window
Claude Haiku 4.5 takes 200K tokens, GLM 5.2 1.05M.

Benchmark by benchmark

First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.

Boards won, where both have a result
CategoryClaude Haiku 4.5BarsGLM 5.2Ties
Coding013
Intelligence02
Agentic work05
Reasoning08
Math03
Knowledge010
Web design08
General119
All shared boards168
CodingLike-for-likeFour of these boards feed the Vibe Coding Index013
Coding benchmarks for Claude Haiku 4.5 and GLM 5.2
BenchmarkClaude Haiku 4.5GLM 5.2Leader
SWE-bench Verified66.6%Not measuredOnly Claude Haiku 4.5 measured
SWE-bench Multilingual64.7%Not measuredOnly Claude Haiku 4.5 measured
LiveBench coding48.1%Not measuredOnly Claude Haiku 4.5 measured
LMArena WebDev13301604GLM 5.2
AA Coding Index43.968.8GLM 5.2
LiveCodeBench (Vals AI)41.2%69.5%GLM 5.2
SWE-bench Verified (Vals AI)66.6%82.8%GLM 5.2
Vibe Code Bench (Vals AI)11.4%64.0%GLM 5.2
Code migration (Vals AI)7.5%37.9%GLM 5.2
ProgramBench (Vals AI)0.0%0.5%GLM 5.2
Terminal-Bench 2.1 (Vals AI)43.8%67.8%GLM 5.2
SkillsBench (Vals AI)Not measured45.1%Only GLM 5.2 measured
SciCodeNot measured51.2%Only GLM 5.2 measured
Terminal-Bench Hard (Artificial Analysis run)Not measured50.8%Only GLM 5.2 measured
Terminal-Bench35.5%Not measuredOnly Claude Haiku 4.5 measured
FrontierCodeNot measured24.5%Only GLM 5.2 measured
WeirdML45.4%70.1%GLM 5.2
ALE-BenchNot measured1047.0Only GLM 5.2 measured
LMArena Text: coding14541487GLM 5.2
LMArena Text: software and IT14391486GLM 5.2
LMArena WebDev: React13261617GLM 5.2
LMArena WebDev: HTML13181565GLM 5.2
Neither model has a result on the other 16 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
Intelligence benchmarks for Claude Haiku 4.5 and GLM 5.2
BenchmarkClaude Haiku 4.5GLM 5.2Leader
AA Intelligence Index16.933.7GLM 5.2
Epoch Capabilities Index142.2151.9GLM 5.2
Agentic workLike-for-likePublic boards, shown as published, kept out of the index05
Agentic work benchmarks for Claude Haiku 4.5 and GLM 5.2
BenchmarkClaude Haiku 4.5GLM 5.2Leader
AA Agentic Index8.038.4GLM 5.2
Design Arena: agents building web appsNot measured1222Only GLM 5.2 measured
Design Arena: agents building full-stack appsNot measured1241Only GLM 5.2 measured
Design Arena: agents building mobile appsNot measured1160Only GLM 5.2 measured
Design Arena: agents building Android appsNot measured1184Only GLM 5.2 measured
Design Arena: agents building Godot gamesNot measured1142Only GLM 5.2 measured
Design Arena: agents building gamesNot measured1206Only GLM 5.2 measured
Design Arena: HTML slidesNot measured1175Only GLM 5.2 measured
Design Arena: PowerPoint slidesNot measured1195Only GLM 5.2 measured
LiveBench research42.2%Not measuredOnly Claude Haiku 4.5 measured
LiveBench browser40.0%Not measuredOnly Claude Haiku 4.5 measured
LiveBench long agentic11.1%Not measuredOnly Claude Haiku 4.5 measured
SREBench (Vals AI)Not measured0.0%Only GLM 5.2 measured
Finance Agent (Vals AI)31.0%49.7%GLM 5.2
Tax Agent Bench (Vals AI)5.9%Not measuredOnly Claude Haiku 4.5 measured
Legal Research Bench (Vals AI)10.6%31.3%GLM 5.2
Excel Modeling Benchmark (Vals AI)22.7%61.5%GLM 5.2
τ²-Bench TelecomNot measured99.1%Only GLM 5.2 measured
APEX-AgentsNot measured45.2%Only GLM 5.2 measured
BALROG31.2%Not measuredOnly Claude Haiku 4.5 measured
DeepResearch Bench45.5%Not measuredOnly Claude Haiku 4.5 measured
PostTrainBenchNot measured31.7%Only GLM 5.2 measured
Vending-Bench 2$459$8,314GLM 5.2
Neither model has a result on the other 16 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index08
Reasoning benchmarks for Claude Haiku 4.5 and GLM 5.2
BenchmarkClaude Haiku 4.5GLM 5.2Leader
GPQA Diamond72.2%85.6%GLM 5.2
LiveBench reasoning61.7%Not measuredOnly Claude Haiku 4.5 measured
MMMU Pro (Vals AI)46.1%Not measuredOnly Claude Haiku 4.5 measured
GPQA Diamond (Artificial Analysis run)Not measured89.5%Only GLM 5.2 measured
Humanity's Last Exam (Artificial Analysis run)Not measured41.1%Only GLM 5.2 measured
GPQA Diamond (Epoch AI run)71.2%91.9%GLM 5.2
ARC-AGI47.7%77.0%GLM 5.2
ARC-AGI-24.0%22.8%GLM 5.2
SimpleBenchNot measured58.8%Only GLM 5.2 measured
Chess puzzles8.0%22.0%GLM 5.2
Mystery game puzzlesNot measured19.0%Only GLM 5.2 measured
EBR-benchNot measured9.5%Only GLM 5.2 measured
LMArena Text: hard prompts14211481GLM 5.2
LMArena Text: expert14431487GLM 5.2
LMArena Text: hard prompts in English14351481GLM 5.2
Neither model has a result on the other 6 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index03
Math benchmarks for Claude Haiku 4.5 and GLM 5.2
BenchmarkClaude Haiku 4.5GLM 5.2Leader
AIME82.7%Not measuredOnly Claude Haiku 4.5 measured
MGSM (Vals AI)92.2%Not measuredOnly Claude Haiku 4.5 measured
OTIS Mock AIME66.7%86.4%GLM 5.2
FrontierMath, tiers 1 to 3Not measured59.2%Only GLM 5.2 measured
FrontierMath, tier 4Not measured29.3%Only GLM 5.2 measured
ProofBenchNot measured35.0%Only GLM 5.2 measured
MATH level 596.4%Not measuredOnly Claude Haiku 4.5 measured
LMArena Text: math13981483GLM 5.2
LMArena Text: mathematical work14301489GLM 5.2
Neither model has a result on the other 4 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index010
Knowledge benchmarks for Claude Haiku 4.5 and GLM 5.2
BenchmarkClaude Haiku 4.5GLM 5.2Leader
MMLU Pro78.7%86.7%GLM 5.2
LegalBench (Vals AI)81.2%84.1%GLM 5.2
CaseLaw (Vals AI)56.5%Not measuredOnly Claude Haiku 4.5 measured
CorpFin (Vals AI)60.6%66.1%GLM 5.2
TaxEval (Vals AI)67.5%73.3%GLM 5.2
MortgageTax (Vals AI)62.2%Not measuredOnly Claude Haiku 4.5 measured
MedQA (Vals AI)79.6%Not measuredOnly Claude Haiku 4.5 measured
MedCode (Vals AI)32.7%40.8%GLM 5.2
SimpleQA Verified13.2%34.2%GLM 5.2
LMArena Text: science14061489GLM 5.2
LMArena Text: medicine and healthcare13941475GLM 5.2
LMArena Text: legal and government14001486GLM 5.2
LMArena Text: business and finance14001451GLM 5.2
Neither model has a result on the other 4 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index08
Web design benchmarks for Claude Haiku 4.5 and GLM 5.2
BenchmarkClaude Haiku 4.5GLM 5.2Leader
Design Arena: websites11221285GLM 5.2
Design Arena: UI components11001296GLM 5.2
Design Arena: game dev10971268GLM 5.2
Design Arena: data viz11241285GLM 5.2
Design Arena: 3D10721288GLM 5.2
Design Arena: all code categories11151288GLM 5.2
Design Arena: SVG10251214GLM 5.2
Design Arena: ASCII art11391190GLM 5.2
LiveBench design59.9%Not measuredOnly Claude Haiku 4.5 measured
GeneralLike-for-likePublic boards, shown as published, kept out of the index119
General benchmarks for Claude Haiku 4.5 and GLM 5.2
BenchmarkClaude Haiku 4.5GLM 5.2Leader
LiveBench data analysis59.3%Not measuredOnly Claude Haiku 4.5 measured
Public Benefits Bench (Vals AI)54.3%Not measuredOnly Claude Haiku 4.5 measured
SAGE (Vals AI)31.8%Not measuredOnly Claude Haiku 4.5 measured
MedScribe (Vals AI)85.2%83.5%Claude Haiku 4.5
Vals Multimodal Index42.9%Not measuredOnly Claude Haiku 4.5 measured
IFBenchNot measured73.3%Only GLM 5.2 measured
Long context reasoning (AA-LCR)Not measured78.3%Only GLM 5.2 measured
DTBench73.6%93.6%GLM 5.2
LMCA30.9%45.8%GLM 5.2
Surface Evolver BenchNot measured55.6%Only GLM 5.2 measured
ExploitBench13.7%Not measuredOnly Claude Haiku 4.5 measured
LMArena Text13971470GLM 5.2
LMArena Text: instruction following14141466GLM 5.2
LMArena Text: creative writing13711463GLM 5.2
LMArena Text: English14141475GLM 5.2
LMArena Text: non-English13771459GLM 5.2
LMArena Text: Chinese14181522GLM 5.2
LMArena Text: German13741470GLM 5.2
LMArena Text: Japanese13381450GLM 5.2
LMArena Text: Korean13461440GLM 5.2
LMArena Text: French14111479GLM 5.2
LMArena Text: Spanish14211478GLM 5.2
LMArena Text: Russian13811468GLM 5.2
LMArena Text: Polish13431456GLM 5.2
LMArena Text: multi-turn14091469GLM 5.2
LMArena Text: longer queries14271480GLM 5.2
LMArena Text: writing and language13861463GLM 5.2
LMArena Text: entertainment and media13691450GLM 5.2
LMArena Document1420Not measuredOnly Claude Haiku 4.5 measured
Neither model has a result on the other 17 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Claude Haiku 4.5 and GLM 5.2
BenchmarkClaude Haiku 4.5GLM 5.2Leader
Neither model has a result on 5 boards here

What each workload costs

Three fixed token mixes at each model's listed price, so per-token rates become a bill.

Chat turn

1K input + 500 output tokens

Claude Haiku 4.5$0.0035Fits in one request
GLM 5.2$0.0035Fits in one request

Claude Haiku 4.5 costs less

Repository review

50K input + 3K output tokens

Claude Haiku 4.5$0.065Fits in one request
GLM 5.2$0.023Fits in one request

GLM 5.2 costs less

Agent loop

200K input + 10K output tokens

Claude Haiku 4.5$0.250Exceeds the context window
GLM 5.2$0.077Fits in one request

GLM 5.2 costs less

Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.

Specification differences

The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.

Specifications of Claude Haiku 4.5 and GLM 5.2
SpecificationAnthropic Claude Haiku 4.5Z.ai GLM 5.2
MakerAnthropicZ.ai
WeightsNo public weights listedOpen weights
ReleasedOctober 15, 2025June 16, 2026
Context window200K1.05M
Input $/Mtok$1.00$0.0341
Output $/Mtok$5.00$7.00
Coding index49.857.7
Independent measurements behind it31
Cost per solved task$0.56$0.86
SourceOpenRouter listingOpenRouter listing

Questions

Is Claude Haiku 4.5 better than GLM 5.2 for coding?

Pick GLM 5.2 for coding: it scores 57.7 on the Vibe Coding Index against 49.8 for Claude Haiku 4.5, though on 1 benchmark against 3. Pick Claude Haiku 4.5 when output tokens cost 1.4x less, or when a solved task costs $0.56 against $0.86.

Which is cheaper, Claude Haiku 4.5 or GLM 5.2?

Claude Haiku 4.5 costs $1.00 per million input tokens and $5.00 per million output tokens; GLM 5.2 costs $0.0341 and $7.00. Per solved SWE-bench Verified task, Claude Haiku 4.5 costs $0.56 and GLM 5.2 $0.86.

Which has the bigger context window, Claude Haiku 4.5 or GLM 5.2?

GLM 5.2, with 1.05M tokens against 200K.

How many benchmarks is the Claude Haiku 4.5 vs GLM 5.2 comparison based on?

69 benchmarks have a result for both models as of October 21, 2026. Claude Haiku 4.5's coding index rests on 3 independent benchmarks and GLM 5.2's on 1. Where only one model has a result, the page says so instead of naming a winner.

Compare next