gpt-oss-20b vs o3
Updated October 21, 2026. 57 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Neither model has a coding score yet, so the two cannot be ranked on coding.
gpt-oss-20b has the edge where output tokens cost 53.3x less, and where what matters is LiveCodeBench (Vals AI) or LiveCodeBench (Artificial Analysis run), where it leads.
- Shared results
- 57
- gpt-oss-20b only
- 3
- o3 only
- 40
- Like-for-like categories
- 8 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
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
o3
o3 leads, 20.2 against 10.0.
Same board, both measured
Reasoning
GPQA Diamond
o3
o3 leads, 84.1% against 68.9%.
Same board, both measured
Math
AIME
gpt-oss-20b
gpt-oss-20b leads, 86.0% against 85.3%.
Same board, both measured
Knowledge
MMLU Pro
o3
o3 leads, 85.6% against 71.6%.
Same board, both measured
Web design
Design Arena: websites
o3
o3 leads, 1036 against 853.
Same board, both measured
Long documents
Listed context window
o3
o3 takes 200K tokens against 131K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
gpt-oss-20b
gpt-oss-20b $0.0001, o3 $0.0060 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only o3 has the number, $0.571.
Directional only
At a glance
- Coding index
- gpt-oss-20b not benchmarked, o3 not benchmarked.
- Widest category
- Web design, 28.1 points apart on average over 2 shared benchmarks.
- Widest single gap
- Long context reasoning (AA-LCR): gpt-oss-20b 34.7% against o3 74.7%, 40.0 points apart on a 0 to 100 scale.
- Output price
- gpt-oss-20b $0.15 against o3 $8.00 per million output tokens, 53.3x apart.
- Context window
- gpt-oss-20b takes 131K tokens, o3 200K.
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 | o3 | Ties |
|---|---|---|---|---|
| Coding | 0 | 7 | ||
| Intelligence | 0 | 2 | ||
| Agentic work | 0 | 1 | ||
| Reasoning | 0 | 7 | ||
| Math | 3 | 5 | ||
| Knowledge | 0 | 10 | ||
| Web design | 0 | 2 | ||
| General | 0 | 20 | ||
| All shared boards | 3 | 54 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index07
| Benchmark | gpt-oss-20b | o3 | Leader |
|---|---|---|---|
| Aider polyglot | Not measured | 81.3% | Only o3 measured |
| SWE-bench Verified | Not measured | 58.4% | Only o3 measured |
| SWE-bench Multimodal | Not measured | 36.0% | Only o3 measured |
| LiveCodeBench (Vals AI) | 80.4% | 83.9% | o3 |
| LiveCodeBench (Artificial Analysis run) | 77.7% | 80.8% | o3 |
| SciCode | 38.9% | Not measured | Only gpt-oss-20b measured |
| Terminal-Bench Hard (Artificial Analysis run) | 10.6% | 37.1% | o3 |
| Terminal-Bench | 3.4% | Not measured | Only gpt-oss-20b measured |
| GSO-Bench | Not measured | 8.8% | Only o3 measured |
| WeirdML | 40.9% | 52.4% | o3 |
| ALE-Bench | 566.0 | 933.5 | o3 |
| LMArena Text: coding | 1307 | 1407 | o3 |
| LMArena Text: software and IT | 1311 | 1415 | o3 |
| Neither model has a result on the other 25 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | gpt-oss-20b | o3 | Leader |
|---|---|---|---|
| AA Intelligence Index | 10.0 | 20.2 | o3 |
| Epoch Capabilities Index | 137.7 | 146.7 | o3 |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | gpt-oss-20b | o3 | Leader |
|---|---|---|---|
| τ²-Bench Telecom | 60.2% | 80.7% | o3 |
| METR time horizons | Not measured | 65.4% | Only o3 measured |
| GDPval | Not measured | 30.8% | Only o3 measured |
| DeepResearch Bench | Not measured | 45.2% | Only o3 measured |
| OSWorld | Not measured | 23.0% | Only o3 measured |
| Neither model has a result on the other 34 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index07
| Benchmark | gpt-oss-20b | o3 | Leader |
|---|---|---|---|
| GPQA Diamond | 68.9% | 84.1% | o3 |
| MMMU Pro (Vals AI) | Not measured | 80.4% | Only o3 measured |
| GPQA Diamond (Artificial Analysis run) | 68.8% | 82.7% | o3 |
| Humanity's Last Exam (Artificial Analysis run) | 11.0% | 20.1% | o3 |
| GPQA Diamond (Epoch AI run) | 60.8% | 81.8% | o3 |
| ARC-AGI | Not measured | 60.8% | Only o3 measured |
| ARC-AGI-2 | Not measured | 6.5% | Only o3 measured |
| SimpleBench | Not measured | 53.1% | Only o3 measured |
| Chess puzzles | Not measured | 37.0% | Only o3 measured |
| Mystery game puzzles | Not measured | 29.0% | Only o3 measured |
| VPCT | Not measured | 52.0% | Only o3 measured |
| LMArena Text: hard prompts | 1274 | 1402 | o3 |
| LMArena Text: expert | 1258 | 1402 | o3 |
| LMArena Text: hard prompts in English | 1292 | 1411 | o3 |
| LMArena Vision: diagrams | Not measured | 1220 | Only o3 measured |
| LMArena Vision: homework | Not measured | 1234 | Only o3 measured |
| Neither model has a result on the other 5 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index35
| Benchmark | gpt-oss-20b | o3 | Leader |
|---|---|---|---|
| AIME | 86.0% | 85.3% | gpt-oss-20b |
| MATH 500 (Vals AI) | 94.2% | 94.6% | o3 |
| MGSM (Vals AI) | 89.0% | 91.8% | o3 |
| AA Math Index | 89.3 | 88.3 | gpt-oss-20b |
| AIME 2025 (Artificial Analysis run) | 89.3% | 88.3% | gpt-oss-20b |
| OTIS Mock AIME | 65.3% | 84.4% | o3 |
| FrontierMath, tiers 1 to 3 | Not measured | 33.3% | Only o3 measured |
| MATH level 5 | Not measured | 97.8% | Only o3 measured |
| LMArena Text: math | 1317 | 1426 | o3 |
| LMArena Text: mathematical work | 1305 | 1423 | o3 |
| Neither model has a result on the other 3 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index010
| Benchmark | gpt-oss-20b | o3 | Leader |
|---|---|---|---|
| MMLU Pro | 71.6% | 85.6% | o3 |
| LegalBench (Vals AI) | 70.8% | 83.8% | o3 |
| CaseLaw (Vals AI) | 43.8% | Not measured | Only gpt-oss-20b measured |
| CorpFin (Vals AI) | 53.1% | 59.7% | o3 |
| TaxEval (Vals AI) | 63.7% | 74.6% | o3 |
| MortgageTax (Vals AI) | Not measured | 65.7% | Only o3 measured |
| MedQA (Vals AI) | 82.9% | 96.1% | o3 |
| MedCode (Vals AI) | Not measured | 47.3% | Only o3 measured |
| MMLU Pro (Artificial Analysis run) | 74.8% | 85.3% | o3 |
| SimpleQA Verified | Not measured | 49.4% | Only o3 measured |
| GeoBench | Not measured | 74.0% | Only o3 measured |
| LMArena Text: science | 1296 | 1429 | o3 |
| LMArena Text: medicine and healthcare | 1310 | 1448 | o3 |
| LMArena Text: legal and government | 1280 | 1428 | o3 |
| LMArena Text: business and finance | 1284 | 1393 | o3 |
| Neither model has a result on the other 2 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | gpt-oss-20b | o3 | Leader |
|---|---|---|---|
| Design Arena: websites | 853 | 1036 | o3 |
| Design Arena: UI components | Not measured | 1017 | Only o3 measured |
| Design Arena: game dev | Not measured | 1034 | Only o3 measured |
| Design Arena: data viz | 933 | 1200 | o3 |
| Design Arena: all code categories | Not measured | 1020 | Only o3 measured |
| Neither model has a result on the other 4 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index020
| Benchmark | gpt-oss-20b | o3 | Leader |
|---|---|---|---|
| SAGE (Vals AI) | Not measured | 41.8% | Only o3 measured |
| MedScribe (Vals AI) | Not measured | 76.7% | Only o3 measured |
| IFBench | 65.1% | 71.4% | o3 |
| Long context reasoning (AA-LCR) | 34.7% | 74.7% | o3 |
| DTBench | 68.0% | 84.8% | o3 |
| LMCA | 14.5% | 39.7% | o3 |
| Fiction.LiveBench | Not measured | 88.9% | Only o3 measured |
| Creative writing (Lech Mazur) | Not measured | 83.9% | Only o3 measured |
| CL-bench | Not measured | 17.8% | Only o3 measured |
| LMArena Text | 1288 | 1410 | o3 |
| LMArena Text: instruction following | 1236 | 1368 | o3 |
| LMArena Text: creative writing | 1201 | 1359 | o3 |
| LMArena Vision | Not measured | 1214 | Only o3 measured |
| LMArena Text: English | 1303 | 1417 | o3 |
| LMArena Text: non-English | 1269 | 1401 | o3 |
| LMArena Text: Chinese | 1315 | 1437 | o3 |
| LMArena Text: German | 1255 | 1420 | o3 |
| LMArena Text: Japanese | 1244 | 1403 | o3 |
| LMArena Text: Korean | 1236 | 1370 | o3 |
| LMArena Text: French | Not measured | 1430 | Only o3 measured |
| LMArena Text: Spanish | 1267 | 1395 | o3 |
| LMArena Text: Russian | 1278 | 1406 | o3 |
| LMArena Text: Polish | 1252 | 1438 | o3 |
| LMArena Text: multi-turn | 1268 | 1405 | o3 |
| LMArena Text: longer queries | 1250 | 1372 | o3 |
| LMArena Text: writing and language | 1238 | 1372 | o3 |
| LMArena Text: entertainment and media | 1216 | 1373 | o3 |
| LMArena Vision: English | Not measured | 1221 | Only o3 measured |
| LMArena Vision: Chinese | Not measured | 1237 | Only o3 measured |
| LMArena Vision: reading text | Not measured | 1218 | Only o3 measured |
| LMArena Vision: humor | Not measured | 1222 | Only o3 measured |
| LMArena Vision: captioning | Not measured | 1210 | Only o3 measured |
| LMArena Vision: recognizing things | Not measured | 1234 | Only o3 measured |
| LMArena Vision: creative writing | Not measured | 1201 | Only o3 measured |
| Neither model has a result on the other 12 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | gpt-oss-20b | o3 | 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 | OpenAI o3 |
|---|---|---|
| Maker | OpenAI | OpenAI |
| Weights | Open weights | No public weights listed |
| Released | August 5, 2025 | April 16, 2025 |
| Context window | 131K | 200K |
| Input $/Mtok | $0.03 | $2.00 |
| Output $/Mtok | $0.15 | $8.00 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 0 |
| Cost per solved task | Not computable | $0.571 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is gpt-oss-20b better than o3 for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. gpt-oss-20b has the edge where output tokens cost 53.3x less, and where what matters is LiveCodeBench (Vals AI) or LiveCodeBench (Artificial Analysis run), where it leads.
Which is cheaper, gpt-oss-20b or o3?
gpt-oss-20b costs $0.03 per million input tokens and $0.15 per million output tokens; o3 costs $2.00 and $8.00. Cost per solved task is known for o3 only.
Which has the bigger context window, gpt-oss-20b or o3?
o3, with 200K tokens against 131K.
How many benchmarks is the gpt-oss-20b vs o3 comparison based on?
57 benchmarks have a result for both models as of October 21, 2026. gpt-oss-20b's coding index rests on 0 independent benchmarks and o3's on 0. Where only one model has a result, the page says so instead of naming a winner.