Model comparison
GPT-5.3 Codex vs Qwen2.5 72B Instruct
GPT-5.3 Codex is the stronger model overall, scoring 45.8 to 31.9 on the Noometry Index. Qwen2.5 72B Instruct costs 2.0× less per token, which makes it the better buy when GPT-5.3 Codex's lead doesn't matter for your workload.
Last verified . 3 shared benchmarks.
Alibaba (Qwen)Summary
- They share 3 benchmarks with published results for both. GPT-5.3 Codex scores higher in 2 categories and Qwen2.5 72B Instruct in 0 categories; 2 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where GPT-5.3 Codex leads 48.0 to 22.1.
- The biggest single-benchmark swing is WeirdML: 79.3% for GPT-5.3 Codex and 16% for Qwen2.5 72B Instruct.
- Qwen2.5 72B Instruct is cheaper at $1.40 / $5.60 per million input/output tokens, against $1.75 / $14 for GPT-5.3 Codex.
- GPT-5.3 Codex accepts more context: 400K tokens versus 131K.
- Qwen2.5 72B Instruct has downloadable open weights; the other is API-only.
Side by side
| GPT-5.3 Codex | Qwen2.5 72B Instruct | |
|---|---|---|
| Provider | OpenAI | Alibaba (Qwen) |
| Noometry Index | 45.8 | 31.9 |
| Released | 2026-02-05 | 2024-09 |
| Weights | Proprietary | Open |
| Context window | 400K | 131K |
| Max output | 128K | 8K |
| Input $ / M tokens | $1.75 | $1.40 |
| Output $ / M tokens | $14 | $5.60 |
| Results tracked | 8 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.3 Codex leads
GPT-5.3 Codex: 48.6 (#56), Qwen2.5 72B Instruct: 33.2 (#260)
| Benchmark | GPT-5.3 Codex | Qwen2.5 72B Instruct |
|---|---|---|
| WeirdML | 79.3% | 16% |
| SWE-bench Verified | 74.8% | — |
| LMArena WebDev | 1409 | — |
| BigCodeBench Instruct | — | 45.8% |
| LMArena Coding | — | 1292 |
| BigCodeBench Complete | — | 55.9% |
| ALE-Bench | 1,655 | — |
Agentic & Tool Use GPT-5.3 Codex leads
GPT-5.3 Codex: 48.0 (#9), Qwen2.5 72B Instruct: 22.1 (#133)
| Benchmark | GPT-5.3 Codex | Qwen2.5 72B Instruct |
|---|---|---|
| METR Time Horizons | 74.5% | 35.8% |
| Terminal-Bench | 78.4% | — |
| TheAgentCompany | — | 5.7% |
| BALROG | — | 16.2% |
| Vending-Bench 2 | 5,940 | — |
Reasoning Not comparable
GPT-5.3 Codex: —, Qwen2.5 72B Instruct: 22.3 (#199)
| Benchmark | GPT-5.3 Codex | Qwen2.5 72B Instruct |
|---|---|---|
| Epoch Capabilities Index | 156.77 | 129 |
| LMArena Hard Prompts | — | 1271 |
| DTBench | — | 62.9% |
| LMCA | — | 13.4% |
| BIG-Bench Hard | — | 79.8% |
| ForecastBench | — | 57.5 |
| HellaSwag | — | 84.8% |
| PIQA | — | 82.6% |
| WinoGrande | — | 82.3% |
Math Not comparable
GPT-5.3 Codex: —, Qwen2.5 72B Instruct: 19.3 (#287)
| Benchmark | GPT-5.3 Codex | Qwen2.5 72B Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | — | 8.1% |
| Omni-MATH | — | 33% |
| LMArena Math | — | 1283 |
| MATH Level 5 | — | 63.2% |
Knowledge Not comparable
GPT-5.3 Codex: —, Qwen2.5 72B Instruct: 27.0 (#253)
| Benchmark | GPT-5.3 Codex | Qwen2.5 72B Instruct |
|---|---|---|
| GPQA Diamond | — | 49.1% |
| MMLU-Pro | — | 63.1% |
| Confabulations | — | 19.1% |
| GPQA (HELM) | — | 42.6% |
| LMArena Expert | — | 1245 |
| ARC (AI2) Challenge | — | 94.5% |
| MMLU | — | 85.3% |
| TriviaQA | — | 71.9% |
Multilingual Not comparable
GPT-5.3 Codex: —, Qwen2.5 72B Instruct: 41.0 (#213)
| Benchmark | GPT-5.3 Codex | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Non-English | — | 1252 |
| LMArena Chinese | — | 1272 |
| LMArena French | — | 1280 |
| LMArena German | — | 1234 |
| LMArena Japanese | — | 1180 |
| LMArena Korean | — | 1188 |
| LMArena Russian | — | 1264 |
| LMArena Spanish | — | 1256 |
Instruction Following Not comparable
GPT-5.3 Codex: —, Qwen2.5 72B Instruct: 65.5 (#221)
| Benchmark | GPT-5.3 Codex | Qwen2.5 72B Instruct |
|---|---|---|
| IFEval | — | 80.6% |
| LMArena Instruction Following | — | 1254 |
Long Context Not comparable
GPT-5.3 Codex: —, Qwen2.5 72B Instruct: 38.9 (#188)
| Benchmark | GPT-5.3 Codex | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Longer Query | — | 1282 |
Writing & Preference Not comparable
GPT-5.3 Codex: —, Qwen2.5 72B Instruct: 46.7 (#215)
| Benchmark | GPT-5.3 Codex | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Text | — | 1269 |
| LMArena Creative Writing | — | 1221 |
| WildBench | — | 80.2% |
| LMArena Multi-Turn | — | 1272 |
Frequently asked questions
Is GPT-5.3 Codex better than Qwen2.5 72B Instruct?
GPT-5.3 Codex is the stronger model overall, scoring 45.8 to 31.9 on the Noometry Index. Qwen2.5 72B Instruct costs 2.0× less per token, which makes it the better buy when GPT-5.3 Codex's lead doesn't matter for your workload.
Which is cheaper, GPT-5.3 Codex or Qwen2.5 72B Instruct?
Qwen2.5 72B Instruct is cheaper. It lists at $1.40 per million input tokens and $5.60 per million output tokens; GPT-5.3 Codex lists at $1.75 and $14.
Is GPT-5.3 Codex or Qwen2.5 72B Instruct better for coding?
GPT-5.3 Codex scores higher on coding benchmarks: 48.6 versus 33.2 in the Noometry coding category.
Which has the bigger context window?
GPT-5.3 Codex does, with 400K tokens against 131K.
How many benchmarks do GPT-5.3 Codex and Qwen2.5 72B Instruct share?
3 benchmarks have published results for both models. GPT-5.3 Codex has 8 scored results on Noometry and Qwen2.5 72B Instruct has 43.