Model comparison
o4-mini vs Qwen3.8 27B
Qwen3.8 27B is the stronger model overall, scoring 46.0 to 41.6 on the Noometry Index.
Last verified . 24 shared benchmarks.
Alibaba (Qwen)Summary
- They share 24 benchmarks with published results for both. o4-mini scores higher in 3 categories and Qwen3.8 27B in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Qwen3.8 27B leads 41.0 to 24.6.
- The biggest single-benchmark swing is ARC-AGI-2: 6.1% for o4-mini and 42.4% for Qwen3.8 27B.
- Qwen3.8 27B is cheaper at $0.04 / $2.30 per million input/output tokens, against $1.10 / $4.40 for o4-mini.
- Qwen3.8 27B accepts more context: 262K tokens versus 200K.
- Qwen3.8 27B has downloadable open weights; the other is API-only.
Side by side
| o4-mini | Qwen3.8 27B | |
|---|---|---|
| Provider | OpenAI | Alibaba (Qwen) |
| Noometry Index | 41.6 | 46.0 |
| Released | 2025-04-16 | 2026-08-14 |
| Weights | Proprietary | Open |
| Context window | 200K | 262K |
| Max output | 100K | 33K |
| Input $ / M tokens | $1.10 | $0.04 |
| Output $ / M tokens | $4.40 | $2.30 |
| Results tracked | 60 | 31 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.8 27B leads
o4-mini: 40.9 (#127), Qwen3.8 27B: 50.5 (#44)
| Benchmark | o4-mini | Qwen3.8 27B |
|---|---|---|
| LMArena Coding | 1368 | 1482 |
| SWE-bench Verified (bash only) | 45% | — |
| Aider Polyglot | 72% | — |
| LMArena WebDev | — | 1593 |
| SciCode | — | 46.6% |
| GSO | 3.6% | — |
| WeirdML | 52.6% | — |
| CadEval | 62% | — |
| ALE-Bench | 826.17 | — |
| AlgoTune | 1.72 | — |
Agentic & Tool Use Too close to call
o4-mini: 32.6 (#61), Qwen3.8 27B: 32.9 (#57)
| Benchmark | o4-mini | Qwen3.8 27B |
|---|---|---|
| APEX-Agents | — | 47.5% |
| Berkeley Function Calling Leaderboard | 53.2% | — |
| GDPval | 25.3% | — |
| METR Time Horizons | 63.9% | — |
Reasoning Qwen3.8 27B leads
o4-mini: 24.6 (#162), Qwen3.8 27B: 41.0 (#54)
| Benchmark | o4-mini | Qwen3.8 27B |
|---|---|---|
| ARC-AGI-2 | 6.1% | 42.4% |
| ARC-AGI-1 | 58.7% | 87.5% |
| CritPt | 0.6% | 5.4% |
| LMArena Hard Prompts | 1351 | 1460 |
| DTBench | 77.6% | 88% |
| LMCA | 26.5% | 41.4% |
| Epoch Capabilities Index | 145.64 | 149.38 |
| SimpleBench | 38.7% | — |
| Kagi LLM Benchmark | 67.6% | — |
| NYT Connections (extended) | — | 54.5% |
| Chess Puzzles | 26% | — |
| EnigmaEval | 9.2% | — |
| Mystery Game Puzzles | 5% | — |
| Surface Evolver Bench | — | 45% |
| ForecastBench | 61.8 | — |
Math o4-mini leads
o4-mini: 40.8 (#89), Qwen3.8 27B: 37.1 (#161)
| Benchmark | o4-mini | Qwen3.8 27B |
|---|---|---|
| LMArena Math | 1389 | 1456 |
| FrontierMath (Tiers 1-3) | 36.1% | — |
| FrontierMath Tier 4 | 4.9% | — |
| OTIS Mock AIME 2024-2025 | 81.7% | — |
| ProofBench | — | 16% |
| Omni-MATH | 72% | — |
| MATH Level 5 | 97.8% | — |
| FrontierMath (Feb 2025 set) | 24.8% | — |
| FrontierMath Tier 4 (v1) | 6.3% | — |
Knowledge o4-mini leads
o4-mini: 43.6 (#91), Qwen3.8 27B: 41.6 (#109)
| Benchmark | o4-mini | Qwen3.8 27B |
|---|---|---|
| LMArena Expert | 1343 | 1482 |
| GPQA Diamond | 79.6% | — |
| Humanity's Last Exam | 18.1% | — |
| SimpleQA Verified | 19.6% | — |
| MMLU-Pro | 82% | — |
| Confabulations | 15.8% | — |
| Vectara Hallucination Rate | 18.6% | — |
| GPQA (HELM) | 73.5% | — |
Multimodal Qwen3.8 27B leads
o4-mini: 40.2 (#49), Qwen3.8 27B: 41.3 (#37)
| Benchmark | o4-mini | Qwen3.8 27B |
|---|---|---|
| LMArena Vision | 1194 | 1271 |
| GeoBench | 64% | — |
| VPCT | 57.5% | — |
Multilingual Qwen3.8 27B leads
o4-mini: 47.0 (#154), Qwen3.8 27B: 53.7 (#60)
| Benchmark | o4-mini | Qwen3.8 27B |
|---|---|---|
| LMArena Non-English | 1337 | 1430 |
| LMArena Chinese | 1354 | 1504 |
| LMArena French | 1364 | 1465 |
| LMArena German | 1336 | 1438 |
| LMArena Japanese | 1308 | 1384 |
| LMArena Korean | 1312 | 1393 |
| LMArena Russian | 1334 | 1415 |
| LMArena Spanish | 1347 | 1448 |
Instruction Following Too close to call
o4-mini: 75.2 (#68), Qwen3.8 27B: 75.8 (#53)
| Benchmark | o4-mini | Qwen3.8 27B |
|---|---|---|
| LMArena Instruction Following | 1321 | 1439 |
| IFEval | 92.8% | — |
Long Context o4-mini leads
o4-mini: 45.5 (#33), Qwen3.8 27B: 44.3 (#70)
| Benchmark | o4-mini | Qwen3.8 27B |
|---|---|---|
| LMArena Longer Query | 1315 | 1450 |
| Fiction.LiveBench | 77.8% | — |
Writing & Preference Qwen3.8 27B leads
o4-mini: 54.0 (#152), Qwen3.8 27B: 65.8 (#43)
| Benchmark | o4-mini | Qwen3.8 27B |
|---|---|---|
| LMArena Text | 1353 | 1441 |
| LMArena Creative Writing | 1294 | 1384 |
| LMArena Multi-Turn | 1350 | 1441 |
| Short-Story Creative Writing | 75% | — |
| EQ-Bench Creative Writing | — | 1671 |
| WildBench | 85.4% | — |
Frequently asked questions
Is o4-mini better than Qwen3.8 27B?
Qwen3.8 27B is the stronger model overall, scoring 46.0 to 41.6 on the Noometry Index.
Which is cheaper, o4-mini or Qwen3.8 27B?
Qwen3.8 27B is cheaper. It lists at $0.04 per million input tokens and $2.30 per million output tokens; o4-mini lists at $1.10 and $4.40.
Is o4-mini or Qwen3.8 27B better for coding?
Qwen3.8 27B scores higher on coding benchmarks: 50.5 versus 40.9 in the Noometry coding category.
Which has the bigger context window?
Qwen3.8 27B does, with 262K tokens against 200K.
How many benchmarks do o4-mini and Qwen3.8 27B share?
24 benchmarks have published results for both models. o4-mini has 60 scored results on Noometry and Qwen3.8 27B has 31.