Model comparison
Claude Sonnet 4.6 vs GPT-5.4 mini
Claude Sonnet 4.6 is the stronger model overall, scoring 50.1 to 44.8 on the Noometry Index. GPT-5.4 mini costs 3.6× less per token, which makes it the better buy when Claude Sonnet 4.6's lead doesn't matter for your workload.
Last verified . 43 shared benchmarks.
Summary
- They share 43 benchmarks with published results for both. Claude Sonnet 4.6 scores higher in 9 categories and GPT-5.4 mini in 1 category; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Claude Sonnet 4.6 leads 46.1 to 30.3.
- The biggest single-benchmark swing is ARC-AGI-2: 60.4% for Claude Sonnet 4.6 and 18.9% for GPT-5.4 mini.
- GPT-5.4 mini is cheaper at $0.75 / $4.50 per million input/output tokens, against $3 / $15 for Claude Sonnet 4.6.
- Claude Sonnet 4.6 accepts more context: 1M tokens versus 400K.
Side by side
| Claude Sonnet 4.6 | GPT-5.4 mini | |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 50.1 | 44.8 |
| Released | 2026-02-17 | 2026-03-17 |
| Weights | Proprietary | Proprietary |
| Context window | 1M | 400K |
| Max output | 128K | 128K |
| Input $ / M tokens | $3 | $0.75 |
| Output $ / M tokens | $15 | $4.50 |
| Results tracked | 57 | 46 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude Sonnet 4.6 leads
Claude Sonnet 4.6: 46.0 (#62), GPT-5.4 mini: 45.0 (#67)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 mini |
|---|---|---|
| FrontierCode | 24.3% | 27% |
| LMArena WebDev | 1521 | 1397 |
| SciCode | 46.8% | 49.9% |
| WeirdML | 66.1% | 60.3% |
| LMArena Coding | 1504 | 1438 |
| ALE-Bench | 1,327 | 1,189 |
| SWE-bench Verified | 75.2% | — |
| DeepSWE | 29.9% | — |
Agentic & Tool Use Claude Sonnet 4.6 leads
Claude Sonnet 4.6: 39.1 (#26), GPT-5.4 mini: 29.6 (#74)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 mini |
|---|---|---|
| DeepResearch Bench | 54.9% | 36.3% |
| Terminal-Bench | 53.4% | — |
| APEX-Agents | 43% | — |
| OSWorld 2.0 | 9.3% | — |
| OSWorld | 72.1% | — |
| ExploitBench | 23.6% | — |
| GBAEval | 48.8% | — |
| GDP.pdf | 18% | — |
| LMArena Search | 1221 | — |
| Vending-Bench 2 | 7,204 | — |
Reasoning Claude Sonnet 4.6 leads
Claude Sonnet 4.6: 46.1 (#40), GPT-5.4 mini: 30.3 (#77)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 mini |
|---|---|---|
| ARC-AGI-2 | 60.4% | 18.9% |
| NYT Connections (extended) | 80.9% | 61.8% |
| ARC-AGI-1 | 86.5% | 63.7% |
| CritPt | 3.1% | 10% |
| Chess Puzzles | 13% | 24% |
| Thematic Generalization | 76.3% | 61.7% |
| LMArena Hard Prompts | 1485 | 1424 |
| Mystery Game Puzzles | 16% | 11% |
| DTBench | 89.9% | 80% |
| LMCA | 46.5% | 40.8% |
| Epoch Capabilities Index | 152.25 | 148.83 |
| ForecastBench | 62 | 57 |
| Kagi LLM Benchmark | — | 37.9% |
Math Claude Sonnet 4.6 leads
Claude Sonnet 4.6: 51.9 (#41), GPT-5.4 mini: 45.4 (#63)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 mini |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 85.8% | 88.9% |
| ProofBench | 45% | 21% |
| LMArena Math | 1463 | 1418 |
| FrontierMath (Feb 2025 set) | 32.4% | 28.3% |
| FrontierMath Tier 4 (v1) | 8.3% | 2.1% |
| FrontierMath (Tiers 1-3) | — | 51.2% |
| FrontierMath Tier 4 | — | 9.8% |
Knowledge Too close to call
Claude Sonnet 4.6: 51.6 (#60), GPT-5.4 mini: 51.4 (#62)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 mini |
|---|---|---|
| GPQA Diamond | 87.4% | 86.9% |
| SimpleQA Verified | 35.5% | 29.4% |
| Vectara Hallucination Rate | 10.6% | 5.5% |
| LMArena Expert | 1500 | 1434 |
Multimodal Too close to call
Claude Sonnet 4.6: 38.0 (#61), GPT-5.4 mini: 39.0 (#51)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 mini |
|---|---|---|
| LMArena Vision | 1283 | 1245 |
| Blueprint-Bench 2 | 6.7% | — |
| LMArena Document | 1482 | — |
Multilingual Claude Sonnet 4.6 leads
Claude Sonnet 4.6: 54.6 (#39), GPT-5.4 mini: 52.1 (#94)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 mini |
|---|---|---|
| LMArena Non-English | 1440 | 1405 |
| LMArena Chinese | 1491 | 1446 |
| LMArena French | 1465 | 1440 |
| LMArena German | 1428 | 1407 |
| LMArena Japanese | 1419 | 1375 |
| LMArena Korean | 1411 | 1369 |
| LMArena Russian | 1440 | 1417 |
| LMArena Spanish | 1465 | 1403 |
Instruction Following Claude Sonnet 4.6 leads
Claude Sonnet 4.6: 77.6 (#23), GPT-5.4 mini: 74.3 (#99)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 mini |
|---|---|---|
| LMArena Instruction Following | 1476 | 1405 |
Long Context Claude Sonnet 4.6 leads
Claude Sonnet 4.6: 45.3 (#40), GPT-5.4 mini: 42.9 (#107)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 mini |
|---|---|---|
| LMArena Longer Query | 1479 | 1407 |
Writing & Preference Claude Sonnet 4.6 leads
Claude Sonnet 4.6: 70.4 (#22), GPT-5.4 mini: 64.2 (#56)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 mini |
|---|---|---|
| LMArena Text | 1458 | 1411 |
| LMArena Creative Writing | 1435 | 1370 |
| EQ-Bench Creative Writing | 1810 | 1665 |
| LMArena Multi-Turn | 1464 | 1429 |
| EQ-Bench 4 | 1207 | — |
Frequently asked questions
Is Claude Sonnet 4.6 better than GPT-5.4 mini?
Claude Sonnet 4.6 is the stronger model overall, scoring 50.1 to 44.8 on the Noometry Index. GPT-5.4 mini costs 3.6× less per token, which makes it the better buy when Claude Sonnet 4.6's lead doesn't matter for your workload.
Which is cheaper, Claude Sonnet 4.6 or GPT-5.4 mini?
GPT-5.4 mini is cheaper. It lists at $0.75 per million input tokens and $4.50 per million output tokens; Claude Sonnet 4.6 lists at $3 and $15.
Is Claude Sonnet 4.6 or GPT-5.4 mini better for coding?
Claude Sonnet 4.6 scores higher on coding benchmarks: 46.0 versus 45.0 in the Noometry coding category.
Which has the bigger context window?
Claude Sonnet 4.6 does, with 1M tokens against 400K.
How many benchmarks do Claude Sonnet 4.6 and GPT-5.4 mini share?
43 benchmarks have published results for both models. Claude Sonnet 4.6 has 57 scored results on Noometry and GPT-5.4 mini has 46.