Model comparison
Claude Haiku 4.5 vs GLM-5.2
GLM-5.2 is the stronger model overall, scoring 50.9 to 39.3 on the Noometry Index.
Last verified . 34 shared benchmarks.
Summary
- They share 34 benchmarks with published results for both. Claude Haiku 4.5 scores higher in 1 category and GLM-5.2 in 8 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GLM-5.2 leads 42.1 to 15.1.
- The biggest single-benchmark swing is NYT Connections (extended): 14.3% for Claude Haiku 4.5 and 74.3% for GLM-5.2.
- Claude Haiku 4.5 is cheaper at $1 / $5 per million input/output tokens, against $1.40 / $4.40 for GLM-5.2.
- GLM-5.2 accepts more context: 1M tokens versus 200K.
- GLM-5.2 has downloadable open weights; the other is API-only.
Side by side
| Claude Haiku 4.5 | GLM-5.2 | |
|---|---|---|
| Provider | Anthropic | Z.ai (Zhipu) |
| Noometry Index | 39.3 | 50.9 |
| Released | 2025-10-15 | 2026-06-13 |
| Weights | Proprietary | Open |
| Context window | 200K | 1M |
| Max output | 64K | 131K |
| Input $ / M tokens | $1 | $1.40 |
| Output $ / M tokens | $5 | $4.40 |
| Results tracked | 53 | 50 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GLM-5.2 leads
Claude Haiku 4.5: 43.9 (#71), GLM-5.2: 50.9 (#37)
| Benchmark | Claude Haiku 4.5 | GLM-5.2 |
|---|---|---|
| LMArena WebDev | 1330 | 1604 |
| SciCode | 43.3% | 50.5% |
| WeirdML | 45.4% | 70.1% |
| LMArena Coding | 1454 | 1484 |
| ALE-Bench | 653.48 | 1,047 |
| SWE-bench Verified | — | 78.7% |
| DeepSWE | — | 43.8% |
| FrontierCode | — | 24.5% |
| SWE-bench Verified (bash only) | 66.6% | — |
| SWE-bench Multilingual | 64.7% | — |
Agentic & Tool Use Claude Haiku 4.5 leads
Claude Haiku 4.5: 33.5 (#50), GLM-5.2: 31.6 (#63)
| Benchmark | Claude Haiku 4.5 | GLM-5.2 |
|---|---|---|
| Vending-Bench 2 | 458.89 | 8,314 |
| Terminal-Bench | 35.5% | — |
| Berkeley Function Calling Leaderboard | 68.7% | — |
| τ²-bench Banking | — | 37.1% |
| DeepResearch Bench | 45.5% | — |
| PostTrainBench | — | 31.7% |
| BALROG | 31.2% | — |
| ExploitBench | 13.7% | — |
| GBAEval | — | 0% |
Reasoning GLM-5.2 leads
Claude Haiku 4.5: 15.1 (#310), GLM-5.2: 42.1 (#46)
| Benchmark | Claude Haiku 4.5 | GLM-5.2 |
|---|---|---|
| ARC-AGI-2 | 4% | 22.8% |
| NYT Connections (extended) | 14.3% | 74.3% |
| ARC-AGI-1 | 47.7% | 77% |
| CritPt | 0% | 20.9% |
| Chess Puzzles | 8% | 21% |
| LMArena Hard Prompts | 1420 | 1480 |
| DTBench | 73.6% | 93.6% |
| LMCA | 30.9% | 45.8% |
| Epoch Capabilities Index | 142.42 | 151.76 |
| SimpleBench | — | 58.8% |
| Kagi LLM Benchmark | — | 62.6% |
| EBR-Bench | — | 9.5% |
| Mystery Game Puzzles | — | 19% |
| Surface Evolver Bench | — | 55.6% |
| ForecastBench | 61.4 | — |
Math GLM-5.2 leads
Claude Haiku 4.5: 43.0 (#69), GLM-5.2: 55.4 (#35)
| Benchmark | Claude Haiku 4.5 | GLM-5.2 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.7% | 86.4% |
| LMArena Math | 1395 | 1482 |
| FrontierMath (Tiers 1-3) | — | 59.2% |
| FrontierMath Tier 4 | — | 29.3% |
| MathArena Final-Answer Competitions | — | 67.6% |
| ProofBench | — | 35% |
| Omni-MATH | 56.1% | — |
| MATH Level 5 | 96.4% | — |
| FrontierMath (Feb 2025 set) | 5.9% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge GLM-5.2 leads
Claude Haiku 4.5: 37.6 (#146), GLM-5.2: 57.1 (#36)
| Benchmark | Claude Haiku 4.5 | GLM-5.2 |
|---|---|---|
| GPQA Diamond | 71.2% | 91.9% |
| SimpleQA Verified | 13.2% | 34.2% |
| LMArena Expert | 1443 | 1484 |
| MMLU-Pro | 77.7% | — |
| Vectara Hallucination Rate | 9.8% | — |
| GPQA (HELM) | 60.5% | — |
Multimodal Not comparable
Claude Haiku 4.5: 27.0 (#111), GLM-5.2: —
| Benchmark | Claude Haiku 4.5 | GLM-5.2 |
|---|---|---|
| Blueprint-Bench 2 | 0% | — |
| LMArena Document | 1420 | — |
Multilingual GLM-5.2 leads
Claude Haiku 4.5: 50.1 (#124), GLM-5.2: 56.0 (#25)
| Benchmark | Claude Haiku 4.5 | GLM-5.2 |
|---|---|---|
| LMArena Non-English | 1376 | 1459 |
| LMArena Chinese | 1415 | 1518 |
| LMArena French | 1408 | 1476 |
| LMArena German | 1377 | 1470 |
| LMArena Japanese | 1339 | 1451 |
| LMArena Korean | 1346 | 1445 |
| LMArena Russian | 1380 | 1467 |
| LMArena Spanish | 1420 | 1477 |
Instruction Following GLM-5.2 leads
Claude Haiku 4.5: 71.6 (#144), GLM-5.2: 77.1 (#32)
| Benchmark | Claude Haiku 4.5 | GLM-5.2 |
|---|---|---|
| LMArena Instruction Following | 1414 | 1465 |
| IFEval | 80.1% | — |
Long Context GLM-5.2 leads
Claude Haiku 4.5: 43.6 (#86), GLM-5.2: 45.3 (#43)
| Benchmark | Claude Haiku 4.5 | GLM-5.2 |
|---|---|---|
| LMArena Longer Query | 1427 | 1479 |
Writing & Preference GLM-5.2 leads
Claude Haiku 4.5: 58.1 (#119), GLM-5.2: 70.7 (#20)
| Benchmark | Claude Haiku 4.5 | GLM-5.2 |
|---|---|---|
| LMArena Text | 1396 | 1470 |
| LMArena Creative Writing | 1371 | 1465 |
| EQ-Bench 4 | 1064 | 1222 |
| LMArena Multi-Turn | 1409 | 1468 |
| EQ-Bench Creative Writing | — | 1757 |
| WildBench | 83.9% | — |
Frequently asked questions
Is Claude Haiku 4.5 better than GLM-5.2?
GLM-5.2 is the stronger model overall, scoring 50.9 to 39.3 on the Noometry Index.
Which is cheaper, Claude Haiku 4.5 or GLM-5.2?
Claude Haiku 4.5 is cheaper. It lists at $1 per million input tokens and $5 per million output tokens; GLM-5.2 lists at $1.40 and $4.40.
Is Claude Haiku 4.5 or GLM-5.2 better for coding?
GLM-5.2 scores higher on coding benchmarks: 50.9 versus 43.9 in the Noometry coding category.
Which has the bigger context window?
GLM-5.2 does, with 1M tokens against 200K.
How many benchmarks do Claude Haiku 4.5 and GLM-5.2 share?
34 benchmarks have published results for both models. Claude Haiku 4.5 has 53 scored results on Noometry and GLM-5.2 has 50.