Model comparison
Claude Haiku 4.5 vs GPT-5.2
GPT-5.2 is the stronger model overall, scoring 54.1 to 39.5 on the Noometry Index. Claude Haiku 4.5 costs 2.4× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.
Last verified . 41 shared benchmarks.
Summary
- They share 41 benchmarks with published results for both. Claude Haiku 4.5 scores higher in 0 categories and GPT-5.2 in 10 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.2 leads 50.2 to 15.1.
- The biggest single-benchmark swing is NYT Connections (extended): 14.3% for Claude Haiku 4.5 and 83.6% for GPT-5.2.
- Claude Haiku 4.5 is cheaper at $1 / $5 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- GPT-5.2 accepts more context: 400K tokens versus 200K.
Side by side
| Claude Haiku 4.5 | GPT-5.2 | |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 39.5 | 54.1 |
| Released | 2025-10-15 | 2025-12-11 |
| Weights | Proprietary | Proprietary |
| Context window | 200K | 400K |
| Max output | 64K | 128K |
| Input $ / M tokens | $1 | $1.75 |
| Output $ / M tokens | $5 | $14 |
| Results tracked | 53 | 67 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.2 leads
Claude Haiku 4.5: 44.0 (#78), GPT-5.2: 51.6 (#37)
| Benchmark | Claude Haiku 4.5 | GPT-5.2 |
|---|---|---|
| SWE-bench Verified (bash only) | 66.6% | 72.8% |
| LMArena WebDev | 1330 | 1416 |
| SWE-bench Multilingual | 64.7% | 66.7% |
| WeirdML | 45.4% | 72.2% |
| LMArena Coding | 1453 | 1447 |
| ALE-Bench | 653.48 | 1,294 |
| SWE-bench Verified | — | 73.8% |
| SciCode | 43.3% | — |
| GSO | — | 27.4% |
| AlgoTune | — | 2.05 |
Agentic & Tool Use GPT-5.2 leads
Claude Haiku 4.5: 33.6 (#52), GPT-5.2: 40.2 (#24)
| Benchmark | Claude Haiku 4.5 | GPT-5.2 |
|---|---|---|
| Terminal-Bench | 35.5% | 64.9% |
| Berkeley Function Calling Leaderboard | 68.7% | 55.9% |
| DeepResearch Bench | 45.5% | 41.1% |
| Vending-Bench 2 | 458.89 | 3,591 |
| GDPval | — | 49.7% |
| Remote Labor Index | — | 2.5% |
| τ²-bench Airline | — | 83% |
| τ²-bench Banking | — | 32.2% |
| τ²-bench Retail | — | 81.6% |
| τ²-bench Telecom | — | 89.7% |
| BALROG | 31.2% | — |
| ExploitBench | 13.7% | — |
| LMArena Search | — | 1207 |
| METR Time Horizons | — | 75.3% |
Reasoning GPT-5.2 leads
Claude Haiku 4.5: 15.1 (#320), GPT-5.2: 50.2 (#35)
| Benchmark | Claude Haiku 4.5 | GPT-5.2 |
|---|---|---|
| ARC-AGI-2 | 4% | 52.9% |
| NYT Connections (extended) | 14.3% | 83.6% |
| ARC-AGI-1 | 47.7% | 86.2% |
| Chess Puzzles | 8% | 49% |
| LMArena Hard Prompts | 1420 | 1445 |
| DTBench | 73.6% | 90.9% |
| LMCA | 30.9% | 43.9% |
| Epoch Capabilities Index | 142.41 | 153.45 |
| ForecastBench | 61.4 | 60.1 |
| SimpleBench | — | 45.8% |
| Kagi LLM Benchmark | — | 73.3% |
| CritPt | 0% | — |
| EnigmaEval | — | 10.4% |
| EBR-Bench | — | 23% |
| Mystery Game Puzzles | — | 23% |
Math GPT-5.2 leads
Claude Haiku 4.5: 44.9 (#78), GPT-5.2: 60.0 (#38)
| Benchmark | Claude Haiku 4.5 | GPT-5.2 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.7% | 96.1% |
| LMArena Math | 1396 | 1440 |
| FrontierMath (Feb 2025 set) | 5.9% | 40.7% |
| FrontierMath Tier 4 (v1) | 2.1% | 18.8% |
| FrontierMath (Tiers 1-3) | — | 67.4% |
| FrontierMath Tier 4 | — | 31.7% |
| MathArena Final-Answer Competitions | — | 72% |
| ProofBench | — | 15% |
| Omni-MATH | 56.1% | — |
| MATH Level 5 | 96.4% | — |
Knowledge GPT-5.2 leads
Claude Haiku 4.5: 37.7 (#153), GPT-5.2: 59.3 (#32)
| Benchmark | Claude Haiku 4.5 | GPT-5.2 |
|---|---|---|
| GPQA Diamond | 71.2% | 91.4% |
| SimpleQA Verified | 13.2% | 37.1% |
| Vectara Hallucination Rate | 9.8% | 8.4% |
| LMArena Expert | 1442 | 1445 |
| Humanity's Last Exam | — | 27.8% |
| MMLU-Pro | 77.7% | — |
| GPQA (HELM) | 60.5% | — |
Multimodal GPT-5.2 leads
Claude Haiku 4.5: 26.8 (#118), GPT-5.2: 51.3 (#7)
| Benchmark | Claude Haiku 4.5 | GPT-5.2 |
|---|---|---|
| LMArena Document | 1420 | 1405 |
| LMArena Vision | — | 1268 |
| VPCT | — | 84% |
| Blueprint-Bench 2 | 0% | — |
| Furniture Assembly | — | 38.3% |
Multilingual GPT-5.2 leads
Claude Haiku 4.5: 49.9 (#129), GPT-5.2: 53.4 (#67)
| Benchmark | Claude Haiku 4.5 | GPT-5.2 |
|---|---|---|
| LMArena Non-English | 1377 | 1425 |
| LMArena Chinese | 1417 | 1460 |
| LMArena French | 1408 | 1455 |
| LMArena German | 1375 | 1448 |
| LMArena Japanese | 1339 | 1420 |
| LMArena Korean | 1347 | 1392 |
| LMArena Russian | 1381 | 1440 |
| LMArena Spanish | 1420 | 1433 |
Instruction Following GPT-5.2 leads
Claude Haiku 4.5: 71.4 (#149), GPT-5.2: 74.7 (#89)
| Benchmark | Claude Haiku 4.5 | GPT-5.2 |
|---|---|---|
| LMArena Instruction Following | 1414 | 1417 |
| IFEval | 80.1% | — |
Long Context Too close to call
Claude Haiku 4.5: 43.6 (#92), GPT-5.2: 44.0 (#78)
| Benchmark | Claude Haiku 4.5 | GPT-5.2 |
|---|---|---|
| LMArena Longer Query | 1427 | 1428 |
| CL-bench | — | 18.2% |
Writing & Preference GPT-5.2 leads
Claude Haiku 4.5: 57.9 (#123), GPT-5.2: 66.8 (#32)
| Benchmark | Claude Haiku 4.5 | GPT-5.2 |
|---|---|---|
| LMArena Text | 1396 | 1439 |
| LMArena Creative Writing | 1372 | 1401 |
| LMArena Multi-Turn | 1409 | 1458 |
| EQ-Bench Creative Writing | — | 1703 |
| WildBench | 83.9% | — |
| EQ-Bench 4 | 1064 | — |
Frequently asked questions
Is Claude Haiku 4.5 better than GPT-5.2?
GPT-5.2 is the stronger model overall, scoring 54.1 to 39.5 on the Noometry Index. Claude Haiku 4.5 costs 2.4× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.
Which is cheaper, Claude Haiku 4.5 or GPT-5.2?
Claude Haiku 4.5 is cheaper. It lists at $1 per million input tokens and $5 per million output tokens; GPT-5.2 lists at $1.75 and $14.
Is Claude Haiku 4.5 or GPT-5.2 better for coding?
GPT-5.2 scores higher on coding benchmarks: 51.6 versus 44.0 in the Noometry coding category.
Which has the bigger context window?
GPT-5.2 does, with 400K tokens against 200K.
How many benchmarks do Claude Haiku 4.5 and GPT-5.2 share?
41 benchmarks have published results for both models. Claude Haiku 4.5 has 53 scored results on Noometry and GPT-5.2 has 67.