Model comparison
GPT-5 Mini vs GPT-6 Astra
GPT-6 Astra is the stronger model overall, scoring 70.3 to 41.7 on the Noometry Index. GPT-5 Mini costs 29× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Last verified . 36 shared benchmarks.
Summary
- They share 36 benchmarks with published results for both. GPT-5 Mini scores higher in 0 categories and GPT-6 Astra in 10 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6 Astra leads 85.1 to 24.1.
- The biggest single-benchmark swing is ARC-AGI-2: 4.4% for GPT-5 Mini and 95% for GPT-6 Astra.
- GPT-5 Mini is cheaper at $0.25 / $2 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
- GPT-6 Astra accepts more context: 1.05M tokens versus 400K.
Side by side
| GPT-5 Mini | GPT-6 Astra | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 41.7 | 70.3 |
| Released | 2025-08-07 | 2026-09-03 |
| Weights | Proprietary | Proprietary |
| Context window | 400K | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $0.25 | $10 |
| Output $ / M tokens | $2 | $50 |
| Results tracked | 60 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Astra leads
GPT-5 Mini: 40.0 (#134), GPT-6 Astra: 70.5 (#3)
| Benchmark | GPT-5 Mini | GPT-6 Astra |
|---|---|---|
| SciCode | 39.2% | 56.5% |
| WeirdML | 52.7% | 93.6% |
| LMArena Coding | 1406 | 1488 |
| ALE-Bench | 799.77 | 2,951 |
| SWE-bench Verified | 64.7% | — |
| DeepSWE | — | 74.1% |
| FrontierCode | — | 53.3% |
| SWE-bench Verified (bash only) | 59.8% | — |
| LMArena WebDev | — | 1792 |
| SWE-bench Multilingual | 39.7% | — |
| FrontierSWE | — | 65.5% |
| MirrorCode | — | 46.7% |
| AlgoTune | 1.38 | — |
Agentic & Tool Use GPT-6 Astra leads
GPT-5 Mini: 31.0 (#64), GPT-6 Astra: 52.3 (#3)
| Benchmark | GPT-5 Mini | GPT-6 Astra |
|---|---|---|
| Vending-Bench 2 | -31.18 | 15,515 |
| Terminal-Bench | 34.8% | — |
| APEX-Agents | — | 64.7% |
| Berkeley Function Calling Leaderboard | 55.5% | — |
| Remote Labor Index | — | 20.8% |
| BALROG | — | 68.3% |
Reasoning GPT-6 Astra leads
GPT-5 Mini: 24.1 (#153), GPT-6 Astra: 85.1 (#1)
| Benchmark | GPT-5 Mini | GPT-6 Astra |
|---|---|---|
| ARC-AGI-2 | 4.4% | 95% |
| ARC-AGI-1 | 54.3% | 98.5% |
| CritPt | 0% | 31.7% |
| Chess Puzzles | 30% | 72% |
| LMArena Hard Prompts | 1380 | 1464 |
| Mystery Game Puzzles | 10% | 84% |
| DTBench | 80.5% | 97.3% |
| LMCA | 34.2% | 64.4% |
| Epoch Capabilities Index | 145.51 | 166.6 |
| Kagi LLM Benchmark | 70.3% | — |
| NYT Connections (extended) | — | 98.1% |
| EnigmaEval | 8.2% | — |
| EBR-Bench | — | 76.2% |
| Bench to the Future 3 | — | 0.14 |
| ForecastBench | 61 | — |
Math GPT-6 Astra leads
GPT-5 Mini: 46.3 (#59), GPT-6 Astra: 93.4 (#1)
| Benchmark | GPT-5 Mini | GPT-6 Astra |
|---|---|---|
| FrontierMath (Tiers 1-3) | 46.7% | 93.7% |
| FrontierMath Tier 4 | 12.2% | 97.6% |
| OTIS Mock AIME 2024-2025 | 86.7% | 100% |
| ProofBench | 9% | 99% |
| LMArena Math | 1377 | 1463 |
| Omni-MATH | 72.2% | — |
| MATH Level 5 | 97.8% | — |
| FrontierMath (Feb 2025 set) | 27.2% | — |
| FrontierMath Erdős | — | 2.9% |
| FrontierMath Tier 4 (v1) | 6.3% | — |
Knowledge GPT-6 Astra leads
GPT-5 Mini: 45.5 (#80), GPT-6 Astra: 75.3 (#1)
| Benchmark | GPT-5 Mini | GPT-6 Astra |
|---|---|---|
| GPQA Diamond | 75% | 95.8% |
| Humanity's Last Exam | 19.4% | 54.8% |
| SimpleQA Verified | 21.6% | 75.6% |
| Vectara Hallucination Rate | 12.9% | 8.7% |
| LMArena Expert | 1377 | 1486 |
| MMLU-Pro | 83.5% | — |
| Confabulations | 13.3% | — |
| GPQA (HELM) | 75.6% | — |
Multimodal GPT-6 Astra leads
GPT-5 Mini: 35.0 (#80), GPT-6 Astra: 55.1 (#3)
| Benchmark | GPT-5 Mini | GPT-6 Astra |
|---|---|---|
| LMArena Vision | 1202 | 1280 |
| VPCT | 40.2% | — |
| Blueprint-Bench 2 | — | 49.7% |
| Furniture Assembly | — | 80% |
| LMArena Document | — | 1468 |
Multilingual GPT-6 Astra leads
GPT-5 Mini: 49.1 (#131), GPT-6 Astra: 54.0 (#53)
| Benchmark | GPT-5 Mini | GPT-6 Astra |
|---|---|---|
| LMArena Non-English | 1363 | 1431 |
| LMArena Chinese | 1387 | 1495 |
| LMArena French | 1385 | 1467 |
| LMArena Russian | 1362 | 1434 |
| LMArena Spanish | 1355 | 1412 |
| LMArena German | 1366 | — |
| LMArena Japanese | 1342 | — |
| LMArena Korean | 1306 | — |
Instruction Following Too close to call
GPT-5 Mini: 76.4 (#43), GPT-6 Astra: 76.6 (#37)
| Benchmark | GPT-5 Mini | GPT-6 Astra |
|---|---|---|
| LMArena Instruction Following | 1357 | 1454 |
| IFEval | 92.7% | — |
Long Context GPT-6 Astra leads
GPT-5 Mini: 41.9 (#127), GPT-6 Astra: 44.7 (#57)
| Benchmark | GPT-5 Mini | GPT-6 Astra |
|---|---|---|
| LMArena Longer Query | 1355 | 1460 |
| Fiction.LiveBench | 69.4% | — |
Writing & Preference GPT-6 Astra leads
GPT-5 Mini: 55.4 (#143), GPT-6 Astra: 75.7 (#7)
| Benchmark | GPT-5 Mini | GPT-6 Astra |
|---|---|---|
| LMArena Text | 1373 | 1443 |
| LMArena Creative Writing | 1325 | 1423 |
| EQ-Bench Creative Writing | 1313 | 2173 |
| LMArena Multi-Turn | 1363 | 1457 |
| Short-Story Creative Writing | 83.1% | — |
| WildBench | 85.5% | — |
Frequently asked questions
Is GPT-5 Mini better than GPT-6 Astra?
GPT-6 Astra is the stronger model overall, scoring 70.3 to 41.7 on the Noometry Index. GPT-5 Mini costs 29× 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-5 Mini or GPT-6 Astra?
GPT-5 Mini is cheaper. It lists at $0.25 per million input tokens and $2 per million output tokens; GPT-6 Astra lists at $10 and $50.
Is GPT-5 Mini or GPT-6 Astra better for coding?
GPT-6 Astra scores higher on coding benchmarks: 70.5 versus 40.0 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 400K.
How many benchmarks do GPT-5 Mini and GPT-6 Astra share?
36 benchmarks have published results for both models. GPT-5 Mini has 60 scored results on Noometry and GPT-6 Astra has 51.