gpt-oss-20b vs Qwen2.5 72B Instruct
gpt-oss-20b leads Qwen2.5 72B Instruct on the Artificial Analysis Intelligence Index, 10.0 to 7.7, and gpt-oss-20b costs 2.7x less per output token ($0.15 vs $0.40 per million). Updated October 21, 2026; 42 benchmarks have a result for both.
gpt-oss-20b or Qwen2.5 72B Instruct: which to pick
Neither model has a coding score yet, so the two cannot be ranked on coding.
Pick gpt-oss-20b for output tokens 2.7x cheaper ($0.15 vs $0.40 per million) and a larger context window (131K against 33K tokens).
- Shared results
- 42
- gpt-oss-20b only
- 20
- Qwen2.5 72B Instruct only
- 11
- Like-for-like categories
- 7 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
gpt-oss-20b
gpt-oss-20b leads, 10.0 against 7.7.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only gpt-oss-20b has a GPQA Diamond result, 68.9%.
Directional only
Math
AIME
No call
Only gpt-oss-20b has a AIME result, 86.0%.
Directional only
Knowledge
MMLU Pro
No call
Only gpt-oss-20b has a MMLU Pro result, 71.6%.
Directional only
Web design
Design Arena: websites
No call
Only gpt-oss-20b has a Design Arena: websites result, 852.
Directional only
Long documents
Listed context window
gpt-oss-20b
gpt-oss-20b takes 131K tokens against 33K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
gpt-oss-20b
gpt-oss-20b $0.0001, Qwen2.5 72B Instruct $0.0006 per turn.
List prices
At a glance
- Coding index
- gpt-oss-20b not benchmarked, Qwen2.5 72B Instruct not benchmarked.
- Widest category
- Math, 43.2 points apart on average over 5 shared benchmarks.
- Widest single gap
- AIME 2025 (Artificial Analysis run): gpt-oss-20b 89.3% against Qwen2.5 72B Instruct 14.0%, 75.3 points apart on a 0 to 100 scale.
- Output price
- gpt-oss-20b $0.15 against Qwen2.5 72B Instruct $0.40 per million output tokens, 2.7x apart.
- Context window
- gpt-oss-20b takes 131K tokens, Qwen2.5 72B Instruct 33K.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | gpt-oss-20b | Bars | Qwen2.5 72B Instruct | Ties |
|---|---|---|---|---|
| Coding | 5 | 0 | ||
| Intelligence | 2 | 0 | ||
| Agentic work | 1 | 0 | ||
| Reasoning | 6 | 0 | ||
| Math | 5 | 0 | ||
| Knowledge | 4 | 1 | ||
| General | 13 | 5 | ||
| All shared boards | 36 | 6 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index50
| Benchmark | gpt-oss-20b | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| LiveCodeBench (Vals AI) | 80.4% | Not measured | Only gpt-oss-20b measured |
| LiveCodeBench (Artificial Analysis run) | 77.7% | 27.6% | gpt-oss-20b |
| SciCode | 38.9% | Not measured | Only gpt-oss-20b measured |
| Terminal-Bench Hard (Artificial Analysis run) | 10.6% | 4.5% | gpt-oss-20b |
| Terminal-Bench | 3.4% | Not measured | Only gpt-oss-20b measured |
| WeirdML | 40.9% | 16.0% | gpt-oss-20b |
| ALE-Bench | 566.0 | Not measured | Only gpt-oss-20b measured |
| LMArena Text: coding | 1307 | 1292 | gpt-oss-20b |
| LMArena Text: software and IT | 1311 | 1292 | gpt-oss-20b |
| Neither model has a result on the other 30 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | gpt-oss-20b | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| AA Intelligence Index | 10.0 | 7.7 | gpt-oss-20b |
| Epoch Capabilities Index | 137.7 | 129.0 | gpt-oss-20b |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | gpt-oss-20b | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| τ²-Bench Telecom | 60.2% | 34.5% | gpt-oss-20b |
| METR time horizons | Not measured | 35.8% | Only Qwen2.5 72B Instruct measured |
| BALROG | Not measured | 16.2% | Only Qwen2.5 72B Instruct measured |
| TheAgentCompany | Not measured | 5.7% | Only Qwen2.5 72B Instruct measured |
| BrowseComp-Plus | 50.6% | Not measured | Only gpt-oss-20b measured |
| Neither model has a result on the other 36 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index60
| Benchmark | gpt-oss-20b | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| GPQA Diamond | 68.9% | Not measured | Only gpt-oss-20b measured |
| GPQA Diamond (Artificial Analysis run) | 68.8% | 49.1% | gpt-oss-20b |
| Humanity's Last Exam (Artificial Analysis run) | 11.0% | 3.6% | gpt-oss-20b |
| GPQA Diamond (Epoch AI run) | 60.8% | 49.1% | gpt-oss-20b |
| BIG-Bench Hard | Not measured | 79.8% | Only Qwen2.5 72B Instruct measured |
| LMArena Text: hard prompts | 1274 | 1271 | gpt-oss-20b |
| LMArena Text: expert | 1258 | 1245 | gpt-oss-20b |
| LMArena Text: hard prompts in English | 1292 | 1279 | gpt-oss-20b |
| Kagi LLM Benchmark | 53.2% | Not measured | Only gpt-oss-20b measured |
| Neither model has a result on the other 13 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index50
| Benchmark | gpt-oss-20b | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| AIME | 86.0% | Not measured | Only gpt-oss-20b measured |
| MATH 500 (Vals AI) | 94.2% | Not measured | Only gpt-oss-20b measured |
| MGSM (Vals AI) | 89.0% | Not measured | Only gpt-oss-20b measured |
| AA Math Index | 89.3 | 14.0 | gpt-oss-20b |
| AIME 2025 (Artificial Analysis run) | 89.3% | 14.0% | gpt-oss-20b |
| OTIS Mock AIME | 65.3% | 8.1% | gpt-oss-20b |
| MATH level 5 | Not measured | 63.2% | Only Qwen2.5 72B Instruct measured |
| LMArena Text: math | 1317 | 1283 | gpt-oss-20b |
| LMArena Text: mathematical work | 1305 | 1290 | gpt-oss-20b |
| Neither model has a result on the other 4 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index41
| Benchmark | gpt-oss-20b | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| MMLU Pro | 71.6% | Not measured | Only gpt-oss-20b measured |
| LegalBench (Vals AI) | 70.8% | Not measured | Only gpt-oss-20b measured |
| CaseLaw (Vals AI) | 43.8% | Not measured | Only gpt-oss-20b measured |
| CorpFin (Vals AI) | 53.1% | Not measured | Only gpt-oss-20b measured |
| TaxEval (Vals AI) | 63.7% | Not measured | Only gpt-oss-20b measured |
| MedQA (Vals AI) | 82.9% | Not measured | Only gpt-oss-20b measured |
| MMLU Pro (Artificial Analysis run) | 74.8% | 72.0% | gpt-oss-20b |
| MMLU | Not measured | 85.3% | Only Qwen2.5 72B Instruct measured |
| AI2 Reasoning Challenge | Not measured | 94.5% | Only Qwen2.5 72B Instruct measured |
| LMArena Text: science | 1296 | 1275 | gpt-oss-20b |
| LMArena Text: medicine and healthcare | 1310 | 1258 | gpt-oss-20b |
| LMArena Text: legal and government | 1280 | 1288 | Qwen2.5 72B Instruct |
| LMArena Text: business and finance | 1284 | 1253 | gpt-oss-20b |
| Neither model has a result on the other 5 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | gpt-oss-20b | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| Design Arena: websites | 852 | Not measured | Only gpt-oss-20b measured |
| Design Arena: data viz | 933 | Not measured | Only gpt-oss-20b measured |
| Neither model has a result on the other 7 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index135
| Benchmark | gpt-oss-20b | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| IFBench | 65.1% | 36.9% | gpt-oss-20b |
| Long context reasoning (AA-LCR) | 34.7% | Not measured | Only gpt-oss-20b measured |
| DTBench | 68.0% | 62.9% | gpt-oss-20b |
| LMCA | 14.5% | 13.4% | gpt-oss-20b |
| HellaSwag | Not measured | 84.8% | Only Qwen2.5 72B Instruct measured |
| WinoGrande | Not measured | 82.3% | Only Qwen2.5 72B Instruct measured |
| PIQA | Not measured | 82.6% | Only Qwen2.5 72B Instruct measured |
| LMArena Text | 1288 | 1269 | gpt-oss-20b |
| LMArena Text: instruction following | 1236 | 1254 | Qwen2.5 72B Instruct |
| LMArena Text: creative writing | 1201 | 1221 | Qwen2.5 72B Instruct |
| LMArena Text: English | 1303 | 1283 | gpt-oss-20b |
| LMArena Text: non-English | 1269 | 1252 | gpt-oss-20b |
| LMArena Text: Chinese | 1315 | 1273 | gpt-oss-20b |
| LMArena Text: German | 1255 | 1234 | gpt-oss-20b |
| LMArena Text: Japanese | 1244 | 1180 | gpt-oss-20b |
| LMArena Text: Korean | 1236 | 1188 | gpt-oss-20b |
| LMArena Text: French | Not measured | 1280 | Only Qwen2.5 72B Instruct measured |
| LMArena Text: Spanish | 1267 | 1255 | gpt-oss-20b |
| LMArena Text: Russian | 1278 | 1264 | gpt-oss-20b |
| LMArena Text: Polish | 1252 | Not measured | Only gpt-oss-20b measured |
| LMArena Text: multi-turn | 1268 | 1272 | Qwen2.5 72B Instruct |
| LMArena Text: longer queries | 1250 | 1281 | Qwen2.5 72B Instruct |
| LMArena Text: writing and language | 1238 | 1247 | Qwen2.5 72B Instruct |
| LMArena Text: entertainment and media | 1216 | 1211 | gpt-oss-20b |
| Neither model has a result on the other 22 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | gpt-oss-20b | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
gpt-oss-20b costs less
Repository review
50K input + 3K output tokens
gpt-oss-20b costs less
Agent loop
200K input + 10K output tokens
gpt-oss-20b costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | OpenAI gpt-oss-20b | Qwen2.5 72B Instruct |
|---|---|---|
| Maker | OpenAI | Qwen |
| Weights | Open weights | Open weights |
| Released | August 5, 2025 | September 19, 2024 |
| Context window | 131K | 33K |
| Input $/Mtok | $0.03 | $0.36 |
| Output $/Mtok | $0.15 | $0.40 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 0 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is gpt-oss-20b better than Qwen2.5 72B Instruct for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Pick gpt-oss-20b for output tokens 2.7x cheaper ($0.15 vs $0.40 per million) and a larger context window (131K against 33K tokens).
Which is cheaper, gpt-oss-20b or Qwen2.5 72B Instruct?
gpt-oss-20b costs $0.03 per million input tokens and $0.15 per million output tokens; Qwen2.5 72B Instruct costs $0.36 and $0.40. Cost per solved task is known for neither of them.
Which has the bigger context window, gpt-oss-20b or Qwen2.5 72B Instruct?
gpt-oss-20b, with 131K tokens against 33K.
How many benchmarks is the gpt-oss-20b vs Qwen2.5 72B Instruct comparison based on?
42 benchmarks have a result for both models as of October 21, 2026. gpt-oss-20b's coding index rests on 0 independent benchmarks and Qwen2.5 72B Instruct's on 0. Where only one model has a result, the page says so instead of naming a winner.