Model comparison
GPT-5.4 vs Qwen2.5-VL 72B Instruct
GPT-5.4 is the stronger model overall, scoring 59.5 to 29.4 on the Noometry Index.
Last verified . 2 shared benchmarks.
Alibaba (Qwen)Summary
- They share 2 benchmarks with published results for both. GPT-5.4 scores higher in 3 categories and Qwen2.5-VL 72B Instruct in 0 categories; 3 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.4 leads 61.7 to 20.0.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 63.8% for GPT-5.4 and 36% for Qwen2.5-VL 72B Instruct.
- Qwen2.5-VL 72B Instruct is cheaper at $2.80 / $8.40 per million input/output tokens, against $2.50 / $15 for GPT-5.4.
- GPT-5.4 accepts more context: 1.05M tokens versus 131K.
- Qwen2.5-VL 72B Instruct has downloadable open weights; the other is API-only.
Side by side
| GPT-5.4 | Qwen2.5-VL 72B Instruct | |
|---|---|---|
| Provider | OpenAI | Alibaba (Qwen) |
| Noometry Index | 59.5 | 29.4 |
| Released | 2026-03-05 | 2024-09 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 131K |
| Max output | 128K | 8K |
| Input $ / M tokens | $2.50 | $2.80 |
| Output $ / M tokens | $15 | $8.40 |
| Results tracked | 68 | 6 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Not comparable
GPT-5.4: 52.8 (#28), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-5.4 | Qwen2.5-VL 72B Instruct |
|---|---|---|
| SWE-bench Verified | 76.9% | — |
| DeepSWE | 51.8% | — |
| LMArena WebDev | 1465 | — |
| SciCode | 56.6% | — |
| GSO | 31.4% | — |
| WeirdML | 77.7% | — |
| LMArena Coding | 1497 | — |
| MirrorCode | 15.6% | — |
| ALE-Bench | 1,607 | — |
| AlgoTune | 1.85 | — |
Agentic & Tool Use GPT-5.4 leads
GPT-5.4: 46.5 (#12), Qwen2.5-VL 72B Instruct: 18.4 (#136)
| Benchmark | GPT-5.4 | Qwen2.5-VL 72B Instruct |
|---|---|---|
| Terminal-Bench | 81.8% | — |
| APEX-Agents | 52.4% | — |
| τ²-bench Banking | 39.4% | — |
| DeepResearch Bench | 35.1% | — |
| OSWorld | — | 5% |
| PostTrainBench | 19% | — |
| GBAEval | 45.1% | — |
| LMArena Search | 1197 | — |
| METR Time Horizons | 74.3% | — |
| Vending-Bench 2 | 6,144 | — |
Reasoning GPT-5.4 leads
GPT-5.4: 61.7 (#17), Qwen2.5-VL 72B Instruct: 20.0 (#228)
| Benchmark | GPT-5.4 | Qwen2.5-VL 72B Instruct |
|---|---|---|
| Kagi LLM Benchmark | 63.8% | 36% |
| ARC-AGI-2 | 74% | — |
| NYT Connections (extended) | 91.3% | — |
| ARC-AGI-1 | 93.7% | — |
| CritPt | 23.4% | — |
| Chess Puzzles | 44% | — |
| EnigmaEval | 16% | — |
| Thematic Generalization | 80% | — |
| EBR-Bench | 25.4% | — |
| LMArena Hard Prompts | 1485 | — |
| Mystery Game Puzzles | 37% | — |
| DTBench | 94.4% | — |
| LMCA | 52% | — |
| Epoch Capabilities Index | 156.92 | — |
| ForecastBench | 59.5 | — |
Math Not comparable
GPT-5.4: 73.3 (#14), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-5.4 | Qwen2.5-VL 72B Instruct |
|---|---|---|
| FrontierMath (Tiers 1-3) | 78.6% | — |
| FrontierMath Tier 4 | 49% | — |
| MathArena Final-Answer Competitions | 83.1% | — |
| OTIS Mock AIME 2024-2025 | 97.8% | — |
| ProofBench | 56% | — |
| LMArena Math | 1488 | — |
| FrontierMath (Feb 2025 set) | 47.6% | — |
| FrontierMath Tier 4 (v1) | 27.1% | — |
Knowledge Not comparable
GPT-5.4: 65.3 (#13), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-5.4 | Qwen2.5-VL 72B Instruct |
|---|---|---|
| GPQA Diamond | 93.3% | — |
| Humanity's Last Exam | 36.2% | — |
| SimpleQA Verified | 45.1% | — |
| Vectara Hallucination Rate | 7% | — |
| LMArena Expert | 1507 | — |
Multimodal GPT-5.4 leads
GPT-5.4: 43.8 (#19), Qwen2.5-VL 72B Instruct: 32.9 (#92)
| Benchmark | GPT-5.4 | Qwen2.5-VL 72B Instruct |
|---|---|---|
| LMArena Vision | 1302 | 1107 |
| Video-MME | — | 73.5% |
| GeoBench | — | 62% |
| Blueprint-Bench 2 | 27.1% | — |
| Furniture Assembly | 37.5% | — |
| LMArena Document | 1471 | — |
| SpatialViz-Bench | — | 33.3% |
Multilingual Not comparable
GPT-5.4: 56.4 (#22), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-5.4 | Qwen2.5-VL 72B Instruct |
|---|---|---|
| LMArena Non-English | 1465 | — |
| LMArena Chinese | 1520 | — |
| LMArena French | 1493 | — |
| LMArena German | 1472 | — |
| LMArena Japanese | 1484 | — |
| LMArena Korean | 1449 | — |
| LMArena Russian | 1480 | — |
| LMArena Spanish | 1453 | — |
Instruction Following Not comparable
GPT-5.4: 77.3 (#25), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-5.4 | Qwen2.5-VL 72B Instruct |
|---|---|---|
| LMArena Instruction Following | 1469 | — |
Long Context Not comparable
GPT-5.4: 50.3 (#8), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-5.4 | Qwen2.5-VL 72B Instruct |
|---|---|---|
| CL-bench | 27.9% | — |
| CL-bench Life | 21.7% | — |
| LMArena Longer Query | 1474 | — |
Writing & Preference Not comparable
GPT-5.4: 72.1 (#18), Qwen2.5-VL 72B Instruct: —
| Benchmark | GPT-5.4 | Qwen2.5-VL 72B Instruct |
|---|---|---|
| LMArena Text | 1469 | — |
| LMArena Creative Writing | 1439 | — |
| EQ-Bench Creative Writing | 1840 | — |
| EQ-Bench 4 | 1272 | — |
| LMArena Multi-Turn | 1482 | — |
Frequently asked questions
Is GPT-5.4 better than Qwen2.5-VL 72B Instruct?
GPT-5.4 is the stronger model overall, scoring 59.5 to 29.4 on the Noometry Index.
Which is cheaper, GPT-5.4 or Qwen2.5-VL 72B Instruct?
Qwen2.5-VL 72B Instruct is cheaper. It lists at $2.80 per million input tokens and $8.40 per million output tokens; GPT-5.4 lists at $2.50 and $15.
Which has the bigger context window?
GPT-5.4 does, with 1.05M tokens against 131K.
How many benchmarks do GPT-5.4 and Qwen2.5-VL 72B Instruct share?
2 benchmarks have published results for both models. GPT-5.4 has 68 scored results on Noometry and Qwen2.5-VL 72B Instruct has 6.