Model comparison
GPT-6 Astra vs gpt-oss-20b
GPT-6 Astra is the stronger model overall, scoring 70.3 to 32.1 on the Noometry Index. gpt-oss-20b costs 556× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Last verified . 24 shared benchmarks.
Summary
- They share 24 benchmarks with published results for both. GPT-6 Astra scores higher in 9 categories and gpt-oss-20b in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6 Astra leads 85.1 to 18.8.
- The biggest single-benchmark swing is Chess Puzzles: 72% for GPT-6 Astra and 4% for gpt-oss-20b.
- gpt-oss-20b is cheaper at $0.018 / $0.09 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
- GPT-6 Astra accepts more context: 1.05M tokens versus 131K.
- gpt-oss-20b has downloadable open weights; the other is API-only.
Side by side
| GPT-6 Astra | gpt-oss-20b | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 70.3 | 32.1 |
| Released | 2026-09-03 | 2025-08-05 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 131K |
| Max output | 128K | 16K |
| Input $ / M tokens | $10 | $0.018 |
| Output $ / M tokens | $50 | $0.09 |
| Results tracked | 51 | 34 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Astra leads
GPT-6 Astra: 70.5 (#3), gpt-oss-20b: 37.1 (#185)
| Benchmark | GPT-6 Astra | gpt-oss-20b |
|---|---|---|
| SciCode | 56.5% | 34.4% |
| WeirdML | 93.6% | 40.9% |
| LMArena Coding | 1488 | 1307 |
| ALE-Bench | 2,951 | 566.05 |
| DeepSWE | 74.1% | — |
| FrontierCode | 53.3% | — |
| LMArena WebDev | 1792 | — |
| FrontierSWE | 65.5% | — |
| MirrorCode | 46.7% | — |
Agentic & Tool Use GPT-6 Astra leads
GPT-6 Astra: 52.3 (#3), gpt-oss-20b: 9.2 (#146)
| Benchmark | GPT-6 Astra | gpt-oss-20b |
|---|---|---|
| Terminal-Bench | — | 3.4% |
| APEX-Agents | 64.7% | — |
| Remote Labor Index | 20.8% | — |
| BALROG | 68.3% | — |
| Vending-Bench 2 | 15,515 | — |
Reasoning GPT-6 Astra leads
GPT-6 Astra: 85.1 (#1), gpt-oss-20b: 18.8 (#251)
| Benchmark | GPT-6 Astra | gpt-oss-20b |
|---|---|---|
| CritPt | 31.7% | 1.4% |
| Chess Puzzles | 72% | 4% |
| LMArena Hard Prompts | 1464 | 1274 |
| DTBench | 97.3% | 68% |
| LMCA | 64.4% | 14.5% |
| Epoch Capabilities Index | 166.6 | 137.81 |
| ARC-AGI-2 | 95% | — |
| Kagi LLM Benchmark | — | 53.2% |
| NYT Connections (extended) | 98.1% | — |
| ARC-AGI-1 | 98.5% | — |
| EBR-Bench | 76.2% | — |
| Mystery Game Puzzles | 84% | — |
| Bench to the Future 3 | 0.14 | — |
Math GPT-6 Astra leads
GPT-6 Astra: 93.4 (#1), gpt-oss-20b: 37.6 (#93)
| Benchmark | GPT-6 Astra | gpt-oss-20b |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 100% | 65.3% |
| LMArena Math | 1463 | 1317 |
| FrontierMath (Tiers 1-3) | 93.7% | — |
| FrontierMath Tier 4 | 97.6% | — |
| ProofBench | 99% | — |
| Omni-MATH | — | 56.5% |
| FrontierMath Erdős | 2.9% | — |
Knowledge GPT-6 Astra leads
GPT-6 Astra: 75.3 (#1), gpt-oss-20b: 34.6 (#186)
| Benchmark | GPT-6 Astra | gpt-oss-20b |
|---|---|---|
| GPQA Diamond | 95.8% | 60.8% |
| LMArena Expert | 1486 | 1258 |
| Humanity's Last Exam | 54.8% | — |
| SimpleQA Verified | 75.6% | — |
| MMLU-Pro | — | 74% |
| Vectara Hallucination Rate | 8.7% | — |
| GPQA (HELM) | — | 59.4% |
Multimodal Not comparable
GPT-6 Astra: 55.1 (#3), gpt-oss-20b: —
| Benchmark | GPT-6 Astra | gpt-oss-20b |
|---|---|---|
| LMArena Vision | 1280 | — |
| Blueprint-Bench 2 | 49.7% | — |
| Furniture Assembly | 80% | — |
| LMArena Document | 1468 | — |
Multilingual GPT-6 Astra leads
GPT-6 Astra: 54.0 (#53), gpt-oss-20b: 42.4 (#192)
| Benchmark | GPT-6 Astra | gpt-oss-20b |
|---|---|---|
| LMArena Non-English | 1431 | 1269 |
| LMArena Chinese | 1495 | 1311 |
| LMArena Russian | 1434 | 1279 |
| LMArena Spanish | 1412 | 1267 |
| LMArena French | 1467 | — |
| LMArena German | — | 1249 |
| LMArena Japanese | — | 1245 |
| LMArena Korean | — | 1239 |
Instruction Following GPT-6 Astra leads
GPT-6 Astra: 76.6 (#37), gpt-oss-20b: 62.0 (#235)
| Benchmark | GPT-6 Astra | gpt-oss-20b |
|---|---|---|
| LMArena Instruction Following | 1454 | 1236 |
| IFEval | — | 73.2% |
Long Context GPT-6 Astra leads
GPT-6 Astra: 44.7 (#57), gpt-oss-20b: 37.9 (#203)
| Benchmark | GPT-6 Astra | gpt-oss-20b |
|---|---|---|
| LMArena Longer Query | 1460 | 1251 |
Writing & Preference GPT-6 Astra leads
GPT-6 Astra: 75.7 (#7), gpt-oss-20b: 35.8 (#260)
| Benchmark | GPT-6 Astra | gpt-oss-20b |
|---|---|---|
| LMArena Text | 1443 | 1288 |
| LMArena Creative Writing | 1423 | 1201 |
| EQ-Bench Creative Writing | 2173 | 666 |
| LMArena Multi-Turn | 1457 | 1268 |
| WildBench | — | 73.7% |
Frequently asked questions
Is GPT-6 Astra better than gpt-oss-20b?
GPT-6 Astra is the stronger model overall, scoring 70.3 to 32.1 on the Noometry Index. gpt-oss-20b costs 556× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Which is cheaper, GPT-6 Astra or gpt-oss-20b?
gpt-oss-20b is cheaper. It lists at $0.018 per million input tokens and $0.09 per million output tokens; GPT-6 Astra lists at $10 and $50.
Is GPT-6 Astra or gpt-oss-20b better for coding?
GPT-6 Astra scores higher on coding benchmarks: 70.5 versus 37.1 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 131K.
How many benchmarks do GPT-6 Astra and gpt-oss-20b share?
24 benchmarks have published results for both models. GPT-6 Astra has 51 scored results on Noometry and gpt-oss-20b has 34.