Model comparison
DeepSeek V4 Pro vs MiMo-V2.5
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 43.4 on the Noometry Index. MiMo-V2.5 costs 5.7× less per token, which makes it the better buy when DeepSeek V4 Pro's lead doesn't matter for your workload.
Last verified . 22 shared benchmarks.
DeepSeekSummary
- They share 22 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 8 categories and MiMo-V2.5 in 0 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Pro leads 64.8 to 36.8.
- The biggest single-benchmark swing is ProofBench: 50% for DeepSeek V4 Pro and 16% for MiMo-V2.5.
- MiMo-V2.5 is cheaper at $0.14 / $0.28 per million input/output tokens, against $0.66 / $1.98 for DeepSeek V4 Pro.
- MiMo-V2.5 accepts more context: 1.05M tokens versus 1M.
Side by side
| DeepSeek V4 Pro | MiMo-V2.5 | |
|---|---|---|
| Provider | DeepSeek | Xiaomi |
| Noometry Index | 54.3 | 43.4 |
| Released | 2026-04-24 | 2026-04-22 |
| Weights | Open | Open |
| Context window | 1M | 1.05M |
| Max output | 393K | 131K |
| Input $ / M tokens | $0.66 | $0.14 |
| Output $ / M tokens | $1.98 | $0.28 |
| Results tracked | 48 | 23 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek V4 Pro: 52.4 (#34), MiMo-V2.5: 43.9 (#81)
| Benchmark | DeepSeek V4 Pro | MiMo-V2.5 |
|---|---|---|
| LMArena WebDev | 1582 | 1438 |
| SciCode | 51% | 43.1% |
| LMArena Coding | 1470 | 1469 |
| ALE-Bench | 1,403 | 513.95 |
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
| WeirdML | 66.2% | — |
Agentic & Tool Use Not comparable
DeepSeek V4 Pro: 32.8 (#58), MiMo-V2.5: —
| Benchmark | DeepSeek V4 Pro | MiMo-V2.5 |
|---|---|---|
| APEX-Agents | 47.3% | — |
| Vending-Bench 2 | 3,285 | — |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), MiMo-V2.5: 28.6 (#101)
| Benchmark | DeepSeek V4 Pro | MiMo-V2.5 |
|---|---|---|
| CritPt | 18% | 3.7% |
| LMArena Hard Prompts | 1461 | 1450 |
| ARC-AGI-2 | 61.3% | — |
| Kagi LLM Benchmark | 53.5% | — |
| NYT Connections (extended) | 91.3% | — |
| ARC-AGI-1 | 90.5% | — |
| Chess Puzzles | 47% | — |
| Mystery Game Puzzles | 43% | — |
| DTBench | 93.9% | — |
| LMCA | 45.5% | — |
| Surface Evolver Bench | 40% | — |
| Epoch Capabilities Index | 155.31 | — |
| ForecastBench | 56.1 | — |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.8 (#30), MiMo-V2.5: 36.8 (#163)
| Benchmark | DeepSeek V4 Pro | MiMo-V2.5 |
|---|---|---|
| ProofBench | 50% | 16% |
| LMArena Math | 1455 | 1436 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
| OTIS Mock AIME 2024-2025 | 98.6% | — |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Pro: 59.5 (#31), MiMo-V2.5: 40.8 (#115)
| Benchmark | DeepSeek V4 Pro | MiMo-V2.5 |
|---|---|---|
| LMArena Expert | 1464 | 1460 |
| GPQA Diamond | 91.7% | — |
| SimpleQA Verified | 52.9% | — |
| Vectara Hallucination Rate | 8.6% | — |
Multimodal Not comparable
DeepSeek V4 Pro: —, MiMo-V2.5: 39.8 (#54)
| Benchmark | DeepSeek V4 Pro | MiMo-V2.5 |
|---|---|---|
| LMArena Vision | — | 1247 |
Multilingual DeepSeek V4 Pro leads
DeepSeek V4 Pro: 54.4 (#45), MiMo-V2.5: 51.9 (#99)
| Benchmark | DeepSeek V4 Pro | MiMo-V2.5 |
|---|---|---|
| LMArena Non-English | 1439 | 1404 |
| LMArena Chinese | 1486 | 1468 |
| LMArena French | 1472 | 1447 |
| LMArena German | 1458 | 1421 |
| LMArena Japanese | 1445 | 1306 |
| LMArena Korean | 1447 | 1363 |
| LMArena Russian | 1453 | 1395 |
| LMArena Spanish | 1458 | 1416 |
Instruction Following Too close to call
DeepSeek V4 Pro: 76.1 (#47), MiMo-V2.5: 75.5 (#60)
| Benchmark | DeepSeek V4 Pro | MiMo-V2.5 |
|---|---|---|
| LMArena Instruction Following | 1448 | 1434 |
Long Context Too close to call
DeepSeek V4 Pro: 45.0 (#51), MiMo-V2.5: 44.2 (#73)
| Benchmark | DeepSeek V4 Pro | MiMo-V2.5 |
|---|---|---|
| LMArena Longer Query | 1458 | 1445 |
| CL-bench Life | 13.5% | — |
Writing & Preference DeepSeek V4 Pro leads
DeepSeek V4 Pro: 65.5 (#46), MiMo-V2.5: 61.6 (#86)
| Benchmark | DeepSeek V4 Pro | MiMo-V2.5 |
|---|---|---|
| LMArena Text | 1451 | 1428 |
| LMArena Creative Writing | 1446 | 1393 |
| LMArena Multi-Turn | 1467 | 1445 |
| EQ-Bench Creative Writing | 1553 | — |
| EQ-Bench 4 | 1166 | — |
Frequently asked questions
Is DeepSeek V4 Pro better than MiMo-V2.5?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 43.4 on the Noometry Index. MiMo-V2.5 costs 5.7× less per token, which makes it the better buy when DeepSeek V4 Pro's lead doesn't matter for your workload.
Which is cheaper, DeepSeek V4 Pro or MiMo-V2.5?
MiMo-V2.5 is cheaper. It lists at $0.14 per million input tokens and $0.28 per million output tokens; DeepSeek V4 Pro lists at $0.66 and $1.98.
Is DeepSeek V4 Pro or MiMo-V2.5 better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 43.9 in the Noometry coding category.
Which has the bigger context window?
MiMo-V2.5 does, with 1.05M tokens against 1M.
How many benchmarks do DeepSeek V4 Pro and MiMo-V2.5 share?
22 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and MiMo-V2.5 has 23.