Model comparison
MiMo-V2-Flash vs Qwen3 14B
MiMo-V2-Flash is the stronger model overall, scoring 40.8 to 35.0 on the Noometry Index.
Last verified . 2 shared benchmarks.
Alibaba (Qwen)Summary
- They share 2 benchmarks with published results for both. MiMo-V2-Flash scores higher in 2 categories and Qwen3 14B in 3 categories; 3 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where MiMo-V2-Flash leads 24.1 to 18.1.
- The biggest single-benchmark swing is SciCode: 25.9% for MiMo-V2-Flash and 31.6% for Qwen3 14B.
- MiMo-V2-Flash is cheaper at $0.14 / $0.28 per million input/output tokens, against $0.35 / $1.40 for Qwen3 14B.
- MiMo-V2-Flash accepts more context: 262K tokens versus 131K.
Side by side
| MiMo-V2-Flash | Qwen3 14B | |
|---|---|---|
| Provider | Xiaomi | Alibaba (Qwen) |
| Noometry Index | 40.8 | 35.0 |
| Released | 2025-12-16 | 2025-04 |
| Weights | Open | Open |
| Context window | 262K | 131K |
| Max output | 66K | 8K |
| Input $ / M tokens | $0.14 | $0.35 |
| Output $ / M tokens | $0.28 | $1.40 |
| Results tracked | 21 | 12 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
MiMo-V2-Flash: 35.8 (#205), Qwen3 14B: 36.7 (#191)
| Benchmark | MiMo-V2-Flash | Qwen3 14B |
|---|---|---|
| SciCode | 25.9% | 31.6% |
| LMArena WebDev | 1330 | — |
| LMArena Coding | 1443 | — |
| ALE-Bench | 737.95 | — |
Agentic & Tool Use Not comparable
MiMo-V2-Flash: —, Qwen3 14B: 29.4 (#78)
| Benchmark | MiMo-V2-Flash | Qwen3 14B |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 41% |
Reasoning MiMo-V2-Flash leads
MiMo-V2-Flash: 24.1 (#155), Qwen3 14B: 18.1 (#272)
| Benchmark | MiMo-V2-Flash | Qwen3 14B |
|---|---|---|
| CritPt | 0% | 0% |
| Kagi LLM Benchmark | — | 49.1% |
| Chess Puzzles | — | 4% |
| LMArena Hard Prompts | 1420 | — |
| DTBench | — | 64% |
| LMCA | — | 18.2% |
| Epoch Capabilities Index | — | 138.23 |
Math Qwen3 14B leads
MiMo-V2-Flash: 35.6 (#141), Qwen3 14B: 37.0 (#102)
| Benchmark | MiMo-V2-Flash | Qwen3 14B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | — | 66.4% |
| LMArena Math | 1396 | — |
Knowledge Too close to call
MiMo-V2-Flash: 39.1 (#126), Qwen3 14B: 39.2 (#124)
| Benchmark | MiMo-V2-Flash | Qwen3 14B |
|---|---|---|
| GPQA Diamond | — | 63.8% |
| Vectara Hallucination Rate | — | 5.4% |
| LMArena Expert | 1425 | — |
Multilingual Not comparable
MiMo-V2-Flash: 51.1 (#109), Qwen3 14B: —
| Benchmark | MiMo-V2-Flash | Qwen3 14B |
|---|---|---|
| LMArena Non-English | 1391 | — |
| LMArena Chinese | 1463 | — |
| LMArena French | 1429 | — |
| LMArena German | 1395 | — |
| LMArena Japanese | 1324 | — |
| LMArena Korean | 1358 | — |
| LMArena Russian | 1387 | — |
| LMArena Spanish | 1420 | — |
Instruction Following Not comparable
MiMo-V2-Flash: 73.7 (#115), Qwen3 14B: —
| Benchmark | MiMo-V2-Flash | Qwen3 14B |
|---|---|---|
| LMArena Instruction Following | 1392 | — |
Long Context MiMo-V2-Flash leads
MiMo-V2-Flash: 43.0 (#106), Qwen3 14B: 38.0 (#201)
| Benchmark | MiMo-V2-Flash | Qwen3 14B |
|---|---|---|
| Fiction.LiveBench | — | 62.5% |
| LMArena Longer Query | 1409 | — |
Writing & Preference Not comparable
MiMo-V2-Flash: 59.9 (#101), Qwen3 14B: —
| Benchmark | MiMo-V2-Flash | Qwen3 14B |
|---|---|---|
| LMArena Text | 1411 | — |
| LMArena Creative Writing | 1375 | — |
| LMArena Multi-Turn | 1404 | — |
Frequently asked questions
Is MiMo-V2-Flash better than Qwen3 14B?
MiMo-V2-Flash is the stronger model overall, scoring 40.8 to 35.0 on the Noometry Index.
Which is cheaper, MiMo-V2-Flash or Qwen3 14B?
MiMo-V2-Flash is cheaper. It lists at $0.14 per million input tokens and $0.28 per million output tokens; Qwen3 14B lists at $0.35 and $1.40.
Is MiMo-V2-Flash or Qwen3 14B better for coding?
They score almost the same on coding (35.8 vs 36.7); test both on your own repository before choosing.
Which has the bigger context window?
MiMo-V2-Flash does, with 262K tokens against 131K.
How many benchmarks do MiMo-V2-Flash and Qwen3 14B share?
2 benchmarks have published results for both models. MiMo-V2-Flash has 21 scored results on Noometry and Qwen3 14B has 12.